{"id":1406,"date":"2023-08-29T22:08:14","date_gmt":"2023-08-29T20:08:14","guid":{"rendered":"https:\/\/evaltep.xcreative.cz\/uncategorized\/honba-za-vyssi-navratnosti-stoji-opravdu-za-to\/"},"modified":"2023-09-15T15:43:36","modified_gmt":"2023-09-15T13:43:36","slug":"chasing-higher-response-rate-is-it-really-worth-it","status":"publish","type":"post","link":"https:\/\/evaltep.xcreative.cz\/en\/articles\/chasing-higher-response-rate-is-it-really-worth-it\/","title":{"rendered":"Chasing Higher Response Rate. Is It Really Worth It?"},"content":{"rendered":"\n

Chasing Higher Response Rate. Is It Really Worth It?<\/strong><\/p>\n\n\n\n

Abstrakt<\/strong><\/p>\n\n\n\n

P\u0159\u00edsp\u011bvek obrac\u00ed pozornost k m\u00ed\u0159e n\u00e1vratnosti, je\u017e p\u0159edstavuje v\u00fdznamnou charakteristiku mnoh\u00fdch evalua\u010dn\u00edch a v\u00fdzkumn\u00fdch \u0161et\u0159en\u00ed. P\u0159\u00edsp\u011bvek krom\u011b samotn\u00e9ho vymezen\u00ed pojmu nab\u00edz\u00ed souhrn mo\u017en\u00fdch n\u00e1pravn\u00fdch opat\u0159en\u00ed, pokud n\u00e1vratnost nedosahuje o\u010dek\u00e1van\u00e9 m\u00edry, a shrnuje mo\u017en\u00e9 \u00fapravy designu \u0161et\u0159en\u00ed sm\u011b\u0159uj\u00edc\u00ed ke zv\u00fd\u0161en\u00ed n\u00e1vratnosti. Hlavn\u00ed pozornost se p\u0159itom soust\u0159e\u010fuje na dodate\u010dn\u00e9 v\u00fdzvy k \u00fa\u010dasti na v\u00fdzkumu. J\u00e1drem stat\u011b je podrobn\u00fd rozbor \u00fa\u010dink\u016f dodate\u010dn\u00fdch v\u00fdzev nejen na samotnou m\u00edru n\u00e1vratnosti, ale tak\u00e9 na strukturu zkouman\u00e9ho vzorku formovanou dodate\u010dn\u00fdmi v\u00fdzvami, na vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f, ke kter\u00fdm m\u016f\u017ee doj\u00edt v d\u016fsledku pou\u017eit\u00ed dodate\u010dn\u00fdch v\u00fdzev, a v neposledn\u00ed \u0159ad\u011b tak\u00e9 na charakter kl\u00ed\u010dov\u00fdch zji\u0161t\u011bn\u00ed, je\u017e m\u016f\u017ee b\u00fdt dodate\u010dn\u00fdmi v\u00fdzvami ovlivn\u011bn. V textu jsou jednotliv\u00e9 z\u00e1v\u011bry zalo\u017eeny na empirick\u00fdch datech dokumentuj\u00edc\u00edch spokojenost s informa\u010dn\u00edm syst\u00e9mem BENEFIT7. Data poch\u00e1zej\u00ed z v\u00fdzkumn\u00e9ho \u0161et\u0159en\u00ed, je\u017e provedl Institut evaluac\u00ed a soci\u00e1ln\u00edch anal\u00fdz metodou CAWI.<\/p>\n\n\n\n

Abstract<\/strong><\/p>\n\n\n\n

Text turns the attention to nonresponse that is the one of the key characteristic of both evaluation and research inquiries. Besides the review of most important definitions of nonresponse, an overview of selected approaches for improvement the response rate is presented. Moreover, the possible changes to research design are introduced and discussed. However, the main attention is paid to follow-ups, their usefulness and impact on response rate and bias of the research estimates. Text also demonstrates how the follow-ups can influence the structure of the research sample and how it affects the character of achieved results. Specific attention is paid to answering the question of optimal number of follow-up waves. Presented conclusions are based on empirical data describing the satisfaction with information system BENFIT7; such data were collected by Institute for evaluations and social analyses with the use of CAWI.<\/p>\n\n\n\n

Kl\u00ed\u010dov\u00e1 slova<\/strong><\/p>\n\n\n\n

N\u00e1vratnost, vych\u00fdlen\u00ed, reprezentativita, CAWI, dodate\u010dn\u00e9 v\u00fdzvy<\/p>\n\n\n\n

Keywords<\/strong><\/p>\n\n\n\n

Response rate, bias; representativity, CAWI, follow-ups<\/p>\n\n\n\n

1. Diskuse o n\u00e1vratnosti<\/strong><\/p>\n\n\n\n

N\u00e1vratnost \u010di m\u00edra n\u00e1vratnosti (response rate) p\u0159edstavuje jednu z d\u016fle\u017eit\u00fdch charakteristik v\u00fdzkumn\u00fdch \u0161et\u0159en\u00ed, kter\u00e1 odr\u00e1\u017e\u00ed situaci, kdy nejsou k dispozici \u00fadaje od v\u0161ech subjekt\u016f zahrnut\u00fdch do v\u00fdb\u011brov\u00e9ho souboru. Obvykle se rozli\u0161uje unit nonresponse (tj. v\u00fdpadek n\u00e1vratnosti[1] u cel\u00e9ho p\u0159\u00edpadu) a item nonresponse (v\u00fdpadek n\u00e1vratnosti u jednotliv\u00e9 ot\u00e1zky). Unit nonresponse tedy dokumentuje stav, kdy se n\u011bkter\u00e9 ze subjekt\u016f do v\u00fdzkumu v\u016fbec nezapojily, zat\u00edmco item nonresponse reflektuje skute\u010dnost, kdy n\u011bkte\u0159\u00ed respondenti neodpov\u00ed na danou konkr\u00e9tn\u00ed ot\u00e1zku. V t\u00e9to stati je pozornost v\u011bnov\u00e1na v\u00fdpadku n\u00e1vratnosti u cel\u00e9ho p\u0159\u00edpadu, tedy unit nonresponse.<\/p>\n\n\n\n

Mnoh\u00e9 studie (de Leeuw, de Heer 2002; Atrostic et al. 1999; Groves, Couper 1998) prok\u00e1zaly, \u017ee se m\u00edra n\u00e1vratnosti v posledn\u00edch desetilet\u00edch sni\u017euje. Za hlavn\u00ed d\u016fvody tohoto poklesu jsou pova\u017eov\u00e1ny zejm\u00e9na obavy obyvatel ze zneu\u017eit\u00ed osobn\u00edch \u00fadaj\u016f, ned\u016fv\u011bra osloven\u00fdch v prov\u00e1d\u011bn\u00e1 v\u00fdzkumn\u00e1 \u0161et\u0159en\u00ed a nedostatek \u010dasu na stran\u011b respondent\u016f (Bates, Creighton 2000). Mezi d\u016fvody klesaj\u00edc\u00ed n\u00e1vratnosti lze spat\u0159ovat tak\u00e9 zm\u011bny v obvykle pou\u017e\u00edvan\u00e9m designu prov\u00e1d\u011bn\u00fdch \u0161et\u0159en\u00ed, nebo\u0165 n\u011bkter\u00e9 metody sb\u011bru dat pou\u017e\u00edvan\u00e9 v posledn\u00edch letech se vyzna\u010duj\u00ed ni\u017e\u0161\u00ed m\u00edrou n\u00e1vratnosti ne\u017e jin\u00e9[2]. Ke klesaj\u00edc\u00ed n\u00e1vratnosti m\u016f\u017ee rovn\u011b\u017e p\u0159isp\u00edvat sni\u017euj\u00edc\u00ed se kvalita v\u00fdb\u011brov\u00e9 opory (Groves et al. 2002) dostupn\u00e9 pro \u00fa\u010dely prov\u00e1d\u011bn\u00ed v\u00fdzkumn\u00fdch a evalua\u010dn\u00edch \u0161et\u0159en\u00ed.<\/p>\n\n\n\n

1.1 Podstata probl\u00e9mu a jeho v\u00fdznam<\/strong><\/p>\n\n\n\n

Pro adekv\u00e1tn\u00ed posouzen\u00ed n\u00e1vratnosti je zapot\u0159eb\u00ed odli\u0161it ji od dal\u0161\u00edch souvisej\u00edc\u00edch jev\u016f a nast\u00ednit zp\u016fsob, jak n\u00e1vratnost ve v\u00fdzkumn\u00e9 a evalua\u010dn\u00ed praxi korektn\u011b vykazovat. N\u00e1vratnost nej\u010dast\u011bji p\u0159edstavuje pod\u00edl respondent\u016f, kte\u0159\u00ed se zapojili do v\u00fdzkumu na celkov\u00e9m po\u010dtu platn\u00fdch (eligible) jednotek (Bailar, Lanphier 1978 \u010di Madow et al. 1983), p\u0159i\u010dem\u017e za platnou jednotku lze pova\u017eovat subjekty, kter\u00e9 skute\u010dn\u011b n\u00e1le\u017e\u00ed do definovan\u00e9 c\u00edlov\u00e9 skupiny[3]. P\u0159i posuzov\u00e1n\u00ed n\u00e1vratnosti je v\u0161ak t\u0159eba obr\u00e1tit pozornost tak\u00e9 k tzv. pasivn\u00edm kontakt\u016fm, tedy k jednotk\u00e1m, kter\u00e9 zb\u00fdvaj\u00ed po ode\u010dten\u00ed respondent\u016f z celkov\u00e9ho souboru platn\u00fdch jednotek. Curtin, Presser a Singer (2005) v tomto ohledu rozli\u0161uj\u00ed t\u0159i skupiny. Prvn\u00ed p\u0159edstavuj\u00ed jednotlivci, kte\u0159\u00ed sice jsou platn\u00fdmi jednotkami, ale nap\u0159. v d\u016fsledku nespr\u00e1vn\u00fdch kontaktn\u00edch \u00fadaj\u016f nebyla s t\u011bmito jednotkami nav\u00e1z\u00e1na komunikace (non-contacts). Druhou skupinu p\u0159edstavuj\u00ed ti, kte\u0159\u00ed sice byli osloveni, ale explicitn\u011b vyj\u00e1d\u0159ili svou neochotu na dan\u00e9m v\u00fdzkumu participovat (refusals). T\u0159et\u00ed skupinu pak p\u0159edstavuj\u00ed ostatn\u00ed, kte\u0159\u00ed se v\u00fdzkumu z\u00fa\u010dastnit nemohli, nap\u0159. z d\u016fvodu nemoci, kv\u016fli jazykov\u00e9 bari\u00e9\u0159e \u010di v d\u016fsledku zme\u0161k\u00e1n\u00ed term\u00ednu pro vypln\u011bn\u00ed dotazn\u00edku. Jednotliv\u00e9 zp\u016fsoby v\u00fdpo\u010dtu n\u00e1vratnosti se li\u0161\u00ed zahrnov\u00e1n\u00edm \u010di vyd\u011blov\u00e1n\u00edm jednotliv\u00fdch typ\u016f pasivn\u00edch respondent\u016f, nicm\u00e9n\u011b p\u0159evl\u00e1d\u00e1 shoda na zp\u016fsobu v\u00fdpo\u010dtu n\u00e1vratnosti dle metodiky AAPOR[4]. Z hlediska \u00fa\u010delu tohoto textu m\u00e1 rozli\u0161ov\u00e1n\u00ed jednotliv\u00fdch skupin pasivn\u00edch kontakt\u016f z\u00e1sadn\u00ed v\u00fdznam,
nebo\u0165 se k n\u011bmu v\u00e1\u017ee jeden z d\u016fle\u017eit\u00fdch z\u00e1v\u011br\u016f vypl\u00fdvaj\u00edc\u00edch z proveden\u00fdch anal\u00fdz.<\/p>\n\n\n\n

1.2 Faktory ovliv\u0148uj\u00edc\u00ed n\u00e1vratnost<\/strong><\/p>\n\n\n\n

Z rozli\u0161en\u00ed jednotliv\u00fdch skupin pasivn\u00edch kontakt\u016f lze dovodit, \u017ee n\u00edzk\u00e1 n\u00e1vratnost (za n\u00ed\u017e lze pova\u017eovat 10\u201315 %) je bu\u010f d\u016fsledkem odm\u00edtnut\u00ed ze strany osloven\u00fdch jednotlivc\u016f po\u017eadovan\u00e9 \u00fadaje poskytnout (refusals), nebo je d\u016fsledkem neuskute\u010dn\u011bn\u00e9ho kontaktu s potenci\u00e1ln\u00edmi respondenty (non-contacts). Odm\u00edtnut\u00ed m\u016f\u017ee b\u00fdt ovlivn\u011bno individu\u00e1ln\u00edmi vlastnostmi tazatel\u016f a jejich nedostate\u010dn\u00fdmi komunika\u010dn\u00edmi schopnostmi (pokud dan\u00e9 \u0161et\u0159en\u00ed prov\u00e1d\u011bno metodou face-to-face dotazov\u00e1n\u00ed); v p\u0159\u00edpad\u011b on-line v\u00fdzkum\u016f (CAWI) m\u016f\u017ee b\u00fdt odm\u00edtnut\u00ed vyvol\u00e1no grafickou podobou dotazn\u00edku \u010di nedostate\u010dnou p\u0159ehlednost\u00ed p\u0159\u00edslu\u0161n\u00e9 aplikace (Groves et al. 2002). Vysok\u00e1 m\u00edra odm\u00edtnut\u00ed, a t\u00edm vyvolan\u00e1 n\u00edzk\u00e1 n\u00e1vratnost, m\u016f\u017ee b\u00fdt rovn\u011b\u017e odrazem neznalosti evalu\u00e1tora o specifik\u00e1ch c\u00edlov\u00e9 skupiny nebo m\u016f\u017ee poukazovat na neadekv\u00e1tnost pou\u017eit\u00e9ho v\u00fdzkumn\u00e9ho instrumentu. Mezi hlavn\u00ed d\u016fvody neuskute\u010dn\u011bn\u00e9ho kontaktu pat\u0159\u00ed nespr\u00e1vn\u00e9 kontaktn\u00ed \u00fadaje a chyby p\u0159i doru\u010den\u00ed, kdy je nap\u0159. e-mail za\u0159azen do spamu nebo kdy je doru\u010den\u00fd e-mail adres\u00e1tem ignorov\u00e1n.<\/p>\n\n\n\n

V n\u00e1vaznosti na uveden\u00e9 rozli\u0161en\u00ed odm\u00edtnut\u00ed a neuskute\u010dn\u011bn\u00ed kontaktu n\u00e1sleduje podrobn\u011bj\u0161\u00ed rozbor jejich vlivu na n\u00e1vratnost. Groves a Couper (1998) v t\u00e9to souvislosti uv\u00e1d\u011bj\u00ed faktory, kter\u00e9 m\u00edru odm\u00edtnut\u00ed (a pota\u017emo tak\u00e9 n\u00edzkou n\u00e1vratnost) ovliv\u0148uj\u00ed nejv\u00edce. Rozli\u0161uj\u00ed p\u0159itom vlivy exogenn\u00ed vypl\u00fdvaj\u00edc\u00ed z kontextu, v n\u011bm\u017e je dan\u00e9 \u0161et\u0159en\u00ed realizov\u00e1no a evalu\u00e1tor ho nem\u016f\u017ee nijak ovlivnit, od vliv\u016f endogenn\u00edch, kter\u00e9 p\u0159edstavuj\u00ed ovlivniteln\u00e9 charakteristiky v\u00fdzkumn\u00e9ho uspo\u0159\u00e1d\u00e1n\u00ed. Do prvn\u00ed skupiny lze za\u0159adit zejm\u00e9na vliv soci\u00e1ln\u00edho prost\u0159ed\u00ed, tedy nap\u0159. frekvenci prov\u00e1d\u011bn\u00fdch \u0161et\u0159en\u00ed a sociodemografick\u00e9 a socioekonomick\u00e9 charakteristiky c\u00edlov\u00e9 skupiny (profesn\u00ed za\u0159azen\u00ed respondent\u016f, jejich pozice v zam\u011bstn\u00e1n\u00ed apod.).
Mezi v\u00fdznamn\u00e9 faktory zvy\u0161uj\u00edc\u00ed odm\u00edtnut\u00ed d\u00e1le pat\u0159\u00ed n\u00edzk\u00e1 vn\u00edman\u00e1 relevance dan\u00e9ho \u0161et\u0159en\u00ed, obava ze zneu\u017eit\u00ed poskytnut\u00fdch \u00fadaj\u016f, negativn\u00ed postoj k zadavateli\/realiz\u00e1torovi v\u00fdzkumu \u010di nechu\u0165 vyjad\u0159ovat a sd\u00edlet vlastn\u00ed n\u00e1zory (Lessler, Kalsbeek 1992). Odm\u00edtnut\u00ed m\u016f\u017ee naopak sn\u00ed\u017eit pozitivn\u00ed motivace respondent\u016f \u2013 Groves, Singer a Corning (2000) v t\u00e9to souvislosti upozor\u0148uj\u00ed na vy\u0161\u0161\u00ed pravd\u011bpodobnost zapojen\u00ed jednotlivc\u016f, kte\u0159\u00ed maj\u00ed konkr\u00e9tn\u00ed z\u00e1jem na v\u00fdsledc\u00edch prov\u00e1d\u011bn\u00e9ho \u0161et\u0159en\u00ed, a kte\u0159\u00ed jsou p\u0159esv\u011bd\u010deni o tom, \u017ee v\u00fdsledky dan\u00e9ho v\u00fdzkumu mohou ovlivnit jejich pozici[5].<\/p>\n\n\n\n

Druhou skupinu determinant (tj. endogenn\u00ed vlivy) tvo\u0159\u00ed konkr\u00e9tn\u00ed charakteristiky dan\u00e9ho \u0161et\u0159en\u00ed, kter\u00e9 lze souhrnn\u011b ozna\u010dit jako v\u00fdzkumn\u00fd design[6]. Vzhledem k tomu, \u017ee podrobn\u00fd rozbor t\u011bchto faktor\u016f p\u0159esahuje r\u00e1mec tohoto textu, je pozornost zam\u011b\u0159ena pouze na explikaci v\u00fdznamu metod sb\u011bru dat. Plat\u00ed, \u017ee ka\u017ed\u00e1 ze zvolen\u00fdch metod s sebou nese specifick\u00e9 okolnosti, je\u017e odm\u00edtnut\u00ed a pota\u017emo n\u00e1vratnost ovliv\u0148uj\u00ed. V p\u0159\u00edpad\u011b face-to-face interview \u010di CATI[7] m\u016f\u017ee osloven\u00fd odm\u00edtnout \u00fa\u010dast kv\u016fli nevhodn\u00fdm okolnostem, za kter\u00fdch byl tazatelem osloven (tedy nap\u0159. kv\u016fli \u010dasov\u00e9mu zanepr\u00e1zdn\u011bn\u00ed). P\u0159i CAWI m\u016f\u017ee b\u00fdt d\u016fvodem odm\u00edtnut\u00ed nesrozumitelnost pokyn\u016f \u010di nevhodn\u011b zvolen\u00e1 aplikace pro dotazov\u00e1n\u00ed. Na druh\u00e9 stran\u011b lze odm\u00edtnut\u00ed sn\u00ed\u017eit zd\u016frazn\u011bn\u00edm d\u016fle\u017eitosti v\u00fdzkumu ji\u017e v prvotn\u00ed zpr\u00e1v\u011b o prob\u00edhaj\u00edc\u00edm \u0161et\u0159en\u00ed, otev\u0159enou komunikac\u00ed s potenci\u00e1ln\u00edmi respondenty zahrnuj\u00edc\u00ed mj. z\u0159eteln\u011b formulovan\u00e9 c\u00edle v\u00fdzkumu, pou\u017eit\u00fdmi incentivami \u010di podporou ze strany autority, kterou osloven\u00ed uzn\u00e1vaj\u00ed.<\/p>\n\n\n\n

P\u0159i v\u00fdzkumech face-to-face lze konkr\u00e9tn\u00ed d\u016fvod v\u00fdpadku n\u00e1vratnosti snadno zjistit, nebo\u0165 tazatel\u00e9 mohou zaznamenat, zdali konkr\u00e9tn\u00ed osobu na dan\u00e9 adrese nezastihli, \u010di zdali se osloven\u00fd odm\u00edtl v\u00fdzkumu z\u00fa\u010dastnit. V r\u00e1mci \u0161et\u0159en\u00ed bez tazatel\u016f (tedy zejm. CAWI) v\u0161ak d\u016fvody n\u00edzk\u00e9 n\u00e1vratnosti spolehliv\u011b rozli\u0161it nelze.<\/p>\n\n\n\n

1.3 Dopady n\u00edzk\u00e9 n\u00e1vratnosti a zp\u016fsoby \u0159e\u0161en\u00ed<\/strong><\/p>\n\n\n\n

\u0158ada zadavatel\u016f, mnoz\u00ed u\u017eivatel\u00e9 v\u00fdsledk\u016f proveden\u00fdch \u0161et\u0159en\u00ed i evalu\u00e1tor\u016f samotn\u00fdch posuzuje na z\u00e1klad\u011b dosa\u017een\u00e9 n\u00e1vratnosti kvalitu v\u00fdzkumu a \u00fasp\u011b\u0161nost ter\u00e9nn\u00edch prac\u00ed. Tak\u00e9 diskuse v odborn\u00e9 literatu\u0159e se opakovan\u011b vracej\u00ed k ot\u00e1zce, jak\u00e1 m\u00edra n\u00e1vratnosti je akceptovateln\u00e1 a jak\u00e1 rizika p\u0159edstavuje n\u00edzk\u00e1 n\u00e1vratnost z hlediska p\u0159esnosti v\u00fdzkumn\u00fdch odhad\u016f. Z n\u011bkter\u00fdch studi\u00ed je p\u0159itom patrn\u00e9, \u017ee n\u00edzk\u00e1 n\u00e1vratnost skute\u010dn\u011b m\u016f\u017ee sn\u00ed\u017eit reprezentativitu (Lessler, Kalsbeek 1992) a zv\u00fd\u0161it pravd\u011bpodobnost vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f. Na druh\u00e9 stran\u011b v\u0161ak jin\u00e9 v\u00fdzkumy (mj. Cutrin, Presser a Singer 2005; Keeter et al. 2000; Schouten 2004; Merkle, Edelman 2002) ukazuj\u00ed, \u017ee i \u0161et\u0159en\u00ed s n\u00edzkou n\u00e1vratnost\u00ed mohou b\u00fdt reprezentativn\u00ed a mohou na z\u00e1klad\u011b dosa\u017een\u00fdch v\u00fdsledk\u016f umo\u017e\u0148ovat formulaci z\u00e1v\u011br\u016f o cel\u00e9 populaci. Tyto dvojak\u00e9 poznatky dokazuj\u00ed, \u017ee m\u00edra n\u00e1vratnosti nen\u00ed sama o sob\u011b vhodn\u00fdm indik\u00e1torem ani p\u0159esnosti (resp. nevych\u00fdlenosti) v\u00fdzkumn\u00fdch odhad\u016f, ani kvality proveden\u00ed v\u00fdzkumu. P\u0159ed neadekv\u00e1tn\u00edmi z\u00e1v\u011bry dovozovan\u00fdmi z m\u00edry n\u00e1vratnosti varuj\u00ed tak\u00e9 Groves, Fowler a Couper (2004), kte\u0159\u00ed vyvracej\u00ed \u010dast\u00fd omyl, kdy je m\u00edra n\u00e1vratnosti pova\u017eov\u00e1na za ukazatel kvality v\u00fdzkumu.<\/p>\n\n\n\n

N\u00edzk\u00e1 n\u00e1vratnost tedy nesni\u017euje sama o sob\u011b hodnotu proveden\u00e9ho \u0161et\u0159en\u00ed a vyu\u017eitelnost jeho v\u00fdsledk\u016f. Z\u00e1va\u017en\u00fdm nedostatkem, s n\u00edm\u017e je n\u00edzk\u00e1 n\u00e1vratnost mnohdy nespr\u00e1vn\u011b zam\u011b\u0148ov\u00e1na, je vych\u00fdlen\u00ed (bias)[8], jeho\u017e v\u00fdskyt je p\u0159i n\u00edzk\u00e9 n\u00e1vratnosti pravd\u011bpodobn\u011bj\u0161\u00ed (Biemer, Lyberg 2003; Thomsen et al. 2006). Tedy nikoliv n\u00edzk\u00e1 n\u00e1vratnost per se, n\u00fdbr\u017e vych\u00fdlen\u00ed v\u00fdzkumn\u00e9ho odhadu m\u016f\u017ee v\u00e9st k nespr\u00e1vn\u00fdm z\u00e1v\u011br\u016fm o populaci (Groves, Peytcheva 2008; J\u00e4rnbert, \u0150hrvall 2012). Proto by specifikaci n\u00e1vratnosti m\u011bl doprov\u00e1zet tak\u00e9 \u00fadaj o vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f, nap\u0159. ve form\u011b anal\u00fdzy rozd\u00edl\u016f mezi charakteristikami respondent\u016f a pasivn\u00edch kontakt\u016f. V podm\u00ednk\u00e1ch v\u00fdzkumn\u00e9 a evalua\u010dn\u00ed praxe je tedy \u017e\u00e1douc\u00ed vykazovat nejen samotnou m\u00edru n\u00e1vratnosti, ale je tak\u00e9 vhodn\u00e9 testovat schopnost dan\u00e9ho v\u00fdb\u011brov\u00e9ho souboru reprezentovat z\u00e1kladn\u00ed populaci a vhodn\u00fdm zp\u016fsobem o n\u00ed informovat zadavatele a dal\u0161\u00ed u\u017eivatele.<\/p>\n\n\n\n

Pokud je z proveden\u00e9 anal\u00fdzy patrn\u00e9, \u017ee v\u00fdzkumn\u00e9 odhady jsou skute\u010dn\u011b vych\u00fdleny, lze vzniklou situaci \u0159e\u0161it vyu\u017eit\u00edm n\u011bkter\u00e9ho ze zaveden\u00fdch zp\u016fsob\u016f. Jde p\u0159edev\u0161\u00edm o tyto postupy:<\/p>\n\n\n\n

a) substituce<\/p>\n\n\n\n

V tomto p\u0159\u00edpad\u011b jde o n\u00e1hradu subjekt\u016f, kter\u00e9 se do v\u00fdzkumu nezapojily, jin\u00fdmi jednotkami. V praxi se m\u016f\u017ee jednat o dosb\u011br, tedy o proveden\u00ed sb\u011bru dat v dodate\u010dn\u011b koncipovan\u00e9m v\u00fdb\u011bru kompenzuj\u00edc\u00edm nedostatek prim\u00e1rn\u00edho sb\u011bru \u010di o extenzi c\u00edlov\u00e9 skupiny. V takov\u00e9mto p\u0159\u00edpad\u011b mohou b\u00fdt do vzorku zahrnuti nap\u0159. nejen \u017eadatel\u00e9, kte\u0159\u00ed podali projekt v letech 2012\u20132013, ale tak\u00e9 ti, kte\u0159\u00ed podali sv\u00e9 projektov\u00e9 \u017e\u00e1dosti u\u017e v roce 2011.<\/p>\n\n\n\n

b) korekce zji\u0161t\u011bn\u00ed na z\u00e1klad\u011b odhadu v\u00fdchylky<\/p>\n\n\n\n

Porovn\u00e1n\u00edm zn\u00e1m\u00fdch hodnot pro skupinu respondent\u016f a pasivn\u00edch kontakt\u016f (nap\u0159. pod\u00edl mu\u017e\u016f a \u017een ve v\u00fdb\u011bru oproti pod\u00edlu mu\u017e\u016f a \u017een v z\u00e1kladn\u00ed populaci) lze z\u00edskat \u00fadaje o vych\u00fdlen\u00ed z\u00edskan\u00fdch dat. A to jak o vych\u00fdlen\u00ed vzork\u016f oproti z\u00e1kladn\u00ed populaci, tak tak\u00e9 o vych\u00fdlen\u00ed pasivn\u00edch kontakt\u016f v\u016f\u010di souboru respondent\u016f. Tyto \u00fadaje pak mohou b\u00fdt vyu\u017eity pro korekci ostatn\u00edch zji\u0161t\u011bn\u00ed, kde u\u017e takov\u00e1to mo\u017enost srovn\u00e1n\u00ed nen\u00ed. Nast\u00edn\u011bn\u00fd postup je v\u0161ak pom\u011brn\u011b rizikov\u00fd a lze ho doporu\u010dit jen ve v\u00fdjime\u010dn\u00fdch p\u0159\u00edpadech[9].<\/p>\n\n\n\n

c) kompenzace pomoc\u00ed statistick\u00fdch metod<\/p>\n\n\n\n

V tomto p\u0159\u00edpad\u011b jde o pou\u017eit\u00ed specifick\u00fdch postup\u016f, je\u017e mohou zm\u00edrnit dopady n\u00edzk\u00e9 n\u00e1vratnosti. Nej\u010dast\u011bji pou\u017e\u00edvan\u00fdm postupem je aplikace post-stratifika\u010dn\u00ed v\u00e1hy, kdy je relativn\u00ed zastoupen\u00ed jedn\u00e9 podskupiny respondent\u016f pos\u00edleno \u010di oslabeno tak, aby l\u00e9pe odr\u00e1\u017eelo jej\u00ed zastoupen\u00ed v z\u00e1kladn\u00ed populaci. Takov\u00fdto postup v\u0161ak nach\u00e1z\u00ed uplatn\u011bn\u00ed sp\u00ed\u0161e p\u0159i \u0159e\u0161en\u00ed v\u00fdpadku n\u00e1vratnosti u jednotliv\u00e9 ot\u00e1zky (item nonresponse).<\/p>\n\n\n\n

Vzhledem k tomu, \u017ee ka\u017ed\u00fd z uveden\u00fdch postup\u016f s sebou nese \u0159adu rizik, je prvo\u0159ad\u00fdm z\u00e1jmem vych\u00fdlen\u00ed (resp. n\u00edzk\u00e9 n\u00e1vratnosti, je\u017e m\u016f\u017ee vych\u00fdlen\u00ed zp\u016fsobit) p\u0159edch\u00e1zet. Jednotliv\u00e1 preventivn\u00ed opat\u0159en\u00ed pak mohou b\u00fdt zam\u011b\u0159ena jak na zvy\u0161ov\u00e1n\u00ed pravd\u011bpodobnosti, \u017ee \u017e\u00e1douc\u00ed subjekt bude osloven (tedy, \u017ee mu bude doru\u010dena v\u00fdzva k \u00fa\u010dasti na v\u00fdzkumu), tak tak\u00e9 na posilov\u00e1n\u00ed motivace k zapojen\u00ed do v\u00fdzkumu. Keeter et al. (2000) upozor\u0148uj\u00ed, \u017ee zv\u00fd\u0161en\u00e1 n\u00e1vratnost je \u010dastokr\u00e1t vykoupena mimo\u0159\u00e1dn\u00fdm \u00fasil\u00edm vynalo\u017een\u00fdm b\u011bhem ter\u00e9nn\u00edch prac\u00ed. Konkr\u00e9tn\u00ed opat\u0159en\u00ed sm\u011b\u0159uj\u00edc\u00ed ke zvy\u0161ov\u00e1n\u00ed n\u00e1vratnosti p\u0159itom mohou m\u00edt mnoho r\u016fzn\u00fdch forem, z nich\u017e nejv\u00fdznamn\u011bj\u0161\u00ed jsou podrobn\u011bji pops\u00e1ny v n\u00e1sleduj\u00edc\u00edm textu.<\/p>\n\n\n\n

a) Nulov\u00e1 varianta, kter\u00e1 \u010dastokr\u00e1t spont\u00e1nn\u011b vyplyne ze situace, kdy problematice n\u00e1vratnosti nen\u00ed v\u011bnov\u00e1na zvl\u00e1\u0161tn\u00ed pozornost. Z mnoh\u00fdch evalua\u010dn\u00edch zpr\u00e1v z \u010cR i ze zahrani\u010d\u00ed je patrn\u00e9, \u017ee dosa\u017een\u00e1 n\u00e1vratnost b\u00fdv\u00e1 p\u0159ij\u00edm\u00e1na jako fakt beze snahy, z\u00e1jmu \u010di pot\u0159eby p\u0159ij\u00edmat konkr\u00e9tn\u00ed opat\u0159en\u00ed sm\u011b\u0159uj\u00edc\u00ed k jej\u00edmu zv\u00fd\u0161en\u00ed. V mnoha jin\u00fdch p\u0159\u00edpadech je v\u0161ak nulov\u00e1 varianta v\u00fdsledkem z\u00e1m\u011brn\u00e9ho rozhodnut\u00ed zalo\u017een\u00e9ho nap\u0159. na anal\u00fdze dokazuj\u00edc\u00ed, \u017ee se pasivn\u00ed kontakty neodli\u0161uj\u00ed od respondent\u016f. Existuj\u00ed nav\u00edc okolnosti, za kter\u00fdch m\u016f\u017ee b\u00fdt aplikace postup\u016f pro zv\u00fd\u0161en\u00ed n\u00e1vratnosti pova\u017eov\u00e1na za nekorektn\u00ed a eticky \u010di dokonce pr\u00e1vn\u011b spornou. V praxi m\u016f\u017ee b\u00fdt nap\u0159. vysok\u00fd po\u010det dodate\u010dn\u00fdch v\u00fdzev mnoh\u00fdmi adres\u00e1ty vn\u00edm\u00e1n jako obt\u011b\u017eov\u00e1n\u00ed.<\/p>\n\n\n\n

b) Dal\u0161\u00edm z opat\u0159en\u00ed, je\u017e v t\u00e9to souvislosti zmi\u0148uj\u00ed nap\u0159. Lessler a Kalsbeek (1992), je redesign sb\u011bru dat. Jejich doporu\u010den\u00ed sm\u011b\u0159uje zejm\u00e9na k uplatn\u011bn\u00ed mixed-mode[10], ke zv\u00fd\u0161en\u00ed flexibility \u010dasov\u00e1n\u00ed ter\u00e9nn\u00edch prac\u00ed, k volb\u011b adekv\u00e1tn\u00edch incentiv a k proveden\u00ed speci\u00e1ln\u00edho \u0161kolen\u00ed tazatel\u016f zam\u011b\u0159en\u00e9ho na minimalizaci odm\u00edtnut\u00ed.<\/p>\n\n\n\n

c) \u010cast\u00fdm zp\u016fsobem zvy\u0161ov\u00e1n\u00ed n\u00e1vratnosti je opakovan\u00e1 \u017e\u00e1dost o participaci na v\u00fdzkumu. V p\u0159\u00edpad\u011b face-to-face v\u00fdzkum\u016f m\u00e1 typicky podobu opakovan\u00fdch n\u00e1v\u0161t\u011bv tazatele v m\u00edst\u011b bydli\u0161t\u011b \u010di pracovi\u0161t\u011b respondent\u016f, zat\u00edmco p\u0159i v\u00fdzkumech prov\u00e1d\u011bn\u00fdch metodou CATI jde o opakovan\u00e1 vol\u00e1n\u00ed, nez\u0159\u00eddka p\u0159esahuj\u00edc\u00ed 10 pokus\u016f. V r\u00e1mci CAWI v\u00fdzkum\u016f jsou obvykle pou\u017e\u00edv\u00e1ny dodate\u010dn\u00e9 v\u00fdzvy k vypln\u011bn\u00ed dotazn\u00edku (tzv. follow-ups) ur\u010den\u00e9 subjekt\u016fm ve v\u00fdb\u011brov\u00e9m souboru, kter\u00e9 do stanoven\u00e9ho data dan\u00fd dotazn\u00edk nevyplnily[11].<\/p>\n\n\n\n

S ohledem k zam\u011b\u0159en\u00ed t\u00e9to stat\u011b se dal\u0161\u00ed pozornost soust\u0159e\u010fuje na zp\u016fsoby a formy uplatn\u011bn\u00ed dodate\u010dn\u00fdch v\u00fdzev v r\u00e1mci CAWI \u0161et\u0159en\u00ed. Dodate\u010dn\u00e9 v\u00fdzvy mohou b\u00fdt koncipov\u00e1ny r\u016fzn\u00fdm zp\u016fsobem; m\u016f\u017ee b\u00fdt pro n\u011b pou\u017eit identick\u00fd komunika\u010dn\u00ed kan\u00e1l, kter\u00fdm byl distribuov\u00e1n samotn\u00fd dotazn\u00edk nebo m\u016f\u017ee b\u00fdt z\u00e1m\u011brn\u011b zvolen jin\u00fd zp\u016fsob komunikace (nap\u0159. v r\u00e1mci CAWI v\u00fdzkumu mohou b\u00fdt dodate\u010dn\u00e9 v\u00fdzvy provedeny telefonicky). Jednotliv\u00e1 \u0161et\u0159en\u00ed pou\u017e\u00edvaj\u00edc\u00ed dodate\u010dn\u00e9 v\u00fdzvy se d\u00e1le mohou li\u0161it po\u010dtem vln a jejich \u010dasov\u00fdm rozestupem, nap\u0159. Dillman (1972) v t\u00e9to souvislosti navrhuje na z\u00e1klad\u011b vlastn\u00edch pozorov\u00e1n\u00ed realizovat t\u0159i vlny dodate\u010dn\u00fdch v\u00fdzev n\u00e1sledovan\u00e9 po prvn\u00ed pozv\u00e1nce k participaci na v\u00fdzkumu b\u011bhem sedmi t\u00fddn\u016f. Dodate\u010dn\u00e9 v\u00fdzvy se mohou li\u0161it tak\u00e9 obsahovou n\u00e1pln\u00ed: m\u016f\u017ee se jednat o upozorn\u011bn\u00ed na bl\u00ed\u017e\u00edc\u00ed se term\u00edn (Dohremwend 1970), o ozn\u00e1men\u00ed o prodlou\u017een\u00ed term\u00ednu (Ferber, Sudman 1974) \u010di o urgenci. V\u00fdjime\u010dn\u00e9 nen\u00ed ani opakovan\u00e9 zasl\u00e1n\u00ed dotazn\u00edku \u010di linku k on-line dotazn\u00edku, a dokonce se lze setkat i s dodate\u010dn\u011b rozes\u00edlan\u00fdm uji\u0161t\u011bn\u00edm o pot\u0159ebnosti spolupr\u00e1ce (Gunn, Rhodes 1981).<\/p>\n\n\n\n

Na \u00fa\u010dinnost dodate\u010dn\u00fdch v\u00fdzev a na okolnosti jejich uplatn\u011bn\u00ed ve v\u00fdzkumu ukazuje \u0159ada konkr\u00e9tn\u00edch \u0161et\u0159en\u00ed (mj. Dillman 1991; Mehta, Sivadas 1995; Sheehan, Hoy 1999), p\u0159i\u010dem\u017e mnoh\u00e9 prameny (Groves, Cialdini a Couper 1992) zd\u016fraz\u0148uj\u00ed, \u017ee dodate\u010dn\u00e9 v\u00fdzvy maj\u00ed z\u00e1sadn\u00ed vliv na v\u00fdslednou m\u00edru n\u00e1vratnosti. Lohr (2010) v tomto ohledu upozor\u0148uje, \u017ee tlak na zvy\u0161ov\u00e1n\u00ed velikosti v\u00fdb\u011brov\u00e9ho souboru bez adekv\u00e1tn\u00ed anal\u00fdzy pasivn\u00edch kontakt\u016f a zac\u00edlen\u00ed konkr\u00e9tn\u00edch aktivit na tuto skupinu sice m\u016f\u017ee zv\u00fd\u0161it celkov\u00fd po\u010det p\u0159\u00edpad\u016f, nicm\u00e9n\u011b ponech\u00e1v\u00e1 stranou ot\u00e1zku dopadu dodate\u010dn\u00fdch v\u00fdzev na vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f.<\/p>\n\n\n\n

V p\u0159\u00edpadech, kdy se jednotlivci bezprost\u0159edn\u011b reaguj\u00edc\u00ed na \u00favodn\u00ed osloven\u00ed (promptn\u00ed respondenti) neli\u0161\u00ed od ostatn\u00edch, mohou dodate\u010dn\u00e9 v\u00fdzvy zv\u00fd\u0161it n\u00e1vratnost, ani\u017e by t\u00edm do\u0161lo k vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f. Jin\u00e1 situace v\u0161ak nast\u00e1v\u00e1, pokud se z n\u011bjak\u00e9ho d\u016fvodu promptn\u00ed respondenti ve sv\u00fdch postoj\u00edch a chov\u00e1n\u00ed v\u00fdznamn\u011b li\u0161\u00ed od ostatn\u00edch. Potom dodate\u010dn\u00e9 v\u00fdzvy mohou krom\u011b zv\u00fd\u0161en\u00ed n\u00e1vratnosti tak\u00e9 sn\u00ed\u017eit vych\u00fdlen\u00ed, nebo\u0165 do vzorku budou zahrnuti i ti, kte\u0159\u00ed na \u00favodn\u00ed osloven\u00ed nereagovali. Pova\u017euji v\u0161ak za nutn\u00e9 upozornit, \u017ee mohou nastat i situace, kdy se vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f v d\u016fsledku pou\u017eit\u00ed dodate\u010dn\u00fdch v\u00fdzev naopak zv\u00fd\u0161\u00ed.<\/p>\n\n\n\n

A\u010dkoliv plat\u00ed, \u017ee zvy\u0161ov\u00e1n\u00ed n\u00e1vratnosti je nejlep\u0161\u00edm zp\u016fsobem, jak sn\u00ed\u017eit vych\u00fdlen\u00ed (Brick, Bose 2001), je t\u0159eba m\u00edt na pam\u011bti, \u017ee ka\u017ed\u00e9 z opat\u0159en\u00ed m\u00e1 krom\u011b sv\u00fdch v\u00fdhod tak\u00e9 konkr\u00e9tn\u00ed rizika.<\/p>\n\n\n\n

2. Popis analyzovan\u00fdch dat<\/strong><\/p>\n\n\n\n

V\u00fdsledky prezentovan\u00e9 v t\u00e9to stati vznikly na z\u00e1klad\u011b samostatn\u00e9 d\u00edl\u010d\u00ed \u00falohy vych\u00e1zej\u00edc\u00ed z projektu \u201eStanoven\u00ed metodologie zji\u0161\u0165ov\u00e1n\u00ed hodnot vybran\u00fdch indik\u00e1tor\u016f v\u00fdsledk\u016f pro Opera\u010dn\u00ed program Technick\u00e1 pomoc 2014\u20132020 a zji\u0161t\u011bn\u00ed v\u00fdchoz\u00edch hodnot v roce 2014\u201c realizovan\u00e9ho Institutem evaluac\u00ed a soci\u00e1ln\u00edch anal\u00fdz pro Ministerstvo pro m\u00edstn\u00ed rozvoj v roce 2014. P\u016fvodn\u00edm c\u00edlem citovan\u00e9ho projektu bylo prov\u00e9st \u0161et\u0159en\u00ed ve \u010dty\u0159ech c\u00edlov\u00fdch skupin\u00e1ch, nicm\u00e9n\u011b \u00fadaje prezentovan\u00e9 v tomto \u010dl\u00e1nku se s ohledem k p\u0159edm\u011btu zde prezentovan\u00e9 d\u00edl\u010d\u00ed anal\u00fdzy zam\u011b\u0159uj\u00ed jen na jednu z nich, a to na \u017eadatele\/p\u0159\u00edjemce. Obsahov\u00fdm c\u00edlem projektu bylo hodnocen\u00ed spokojenosti vybran\u00fdch akt\u00e9r\u016f s vymezen\u00fdmi oblastmi; \u0161et\u0159en\u00ed realizovan\u00e9 v c\u00edlov\u00e9 skupin\u011b \u017eadatel\u016f\/p\u0159\u00edjemc\u016f se krom\u011b celkov\u00e9 spokojenosti s informa\u010dn\u00edm syst\u00e9mem zam\u011b\u0159ilo tak\u00e9 na spokojenost s pracovn\u00edm prost\u0159ed\u00edm informa\u010dn\u00edho syst\u00e9mu BENEFIT7, spokojenost s daty obsa\u017een\u00fdmi v syst\u00e9mu a spokojenost s technickou podporou.<\/p>\n\n\n\n

\u0160et\u0159en\u00ed bylo realizov\u00e1no metodou CAWI, kter\u00e1 umo\u017enila respondent\u016fm postupovat p\u0159i vypl\u0148ov\u00e1n\u00ed dotazn\u00edku jejich vlastn\u00edm tempem a nav\u00edc absence tazatele vedla k vy\u0161\u0161\u00ed otev\u0159enosti respondent\u016f a sn\u00ed\u017eila tak v\u00fdskyt tzv. soci\u00e1ln\u00ed desirability. Na druh\u00e9 stran\u011b bylo p\u0159i aplikaci metody CAWI t\u0159eba po\u010d\u00edtat s vy\u0161\u0161\u00edm rizikem v\u00fdb\u011brov\u00e9ho vych\u00fdlen\u00ed v d\u016fsledku samov\u00fdb\u011bru. V dan\u00e9m p\u0159\u00edpad\u011b byly obavy spojov\u00e1ny s vy\u0161\u0161\u00edm zapojen\u00edm mimo\u0159\u00e1dn\u011b nespokojen\u00fdch \u010di naopak velmi spokojen\u00fdch jednotlivc\u016f, na \u00fakor \u017eadatel\u016f\/p\u0159\u00edjemc\u016f s nevyhran\u011bn\u00fdmi \u010di neutr\u00e1ln\u00edmi postoji.<\/p>\n\n\n\n

Celkov\u00e1 doba trv\u00e1n\u00ed ter\u00e9nn\u00edch prac\u00ed \u010dinila 15 dn\u00ed a vlastn\u00ed sb\u011br dat prob\u00edhal ve t\u0159ech vln\u00e1ch. Po \u00favodn\u00edm rozesl\u00e1n\u00ed pozv\u00e1nky k zapojen\u00ed do v\u00fdzkumu byl respondent\u016fm ponech\u00e1n \u010dasov\u00fd prostor v d\u00e9lce \u0161esti dn\u00ed. Po uplynut\u00ed t\u00e9to \u0161estidenn\u00ed lh\u016fty byla t\u011bm, kte\u0159\u00ed nereagovali, zasl\u00e1na opakovan\u00e1 v\u00fdzva k zapojen\u00ed do prov\u00e1d\u011bn\u00e9ho v\u00fdzkumu (prvn\u00ed dodate\u010dn\u00e1 v\u00fdzva) a byla stanovena nov\u00e1 lh\u016fta pro vypln\u011bn\u00ed dotazn\u00edku v d\u00e9lce \u010dty\u0159 dn\u00ed. Pot\u00e9 byla rozesl\u00e1na dal\u0161\u00ed v\u00fdzva k participaci (druh\u00e1 dodate\u010dn\u00e1 v\u00fdzva) a po dal\u0161\u00edch 4 dnech byly linky k on-line dotazn\u00edku deaktivov\u00e1ny a proces sb\u011bru dat byl ukon\u010den.<\/p>\n\n\n\n

2.1 C\u00edl anal\u00fdzy<\/strong><\/p>\n\n\n\n

Meritorn\u00ed c\u00edle anal\u00fdzy prezentovan\u00e9 v t\u00e9to stati jsou dva; prvn\u00ed c\u00edlem je ov\u011b\u0159it reprezentativitu vzorku vznikl\u00e9ho v podm\u00ednk\u00e1ch CAWI v\u00fdzkumu v c\u00edlov\u00e9 skupin\u011b \u017eadatel\u016f\/p\u0159\u00edjemc\u016f. Druh\u00fdm c\u00edlem je prozkoumat efekty pou\u017eit\u00ed dodate\u010dn\u00fdch v\u00fdzev. V r\u00e1mci druh\u00e9ho z c\u00edl\u016f se analytick\u00e1 pozornost zam\u011b\u0159ila na zm\u011bny n\u00e1vratnosti po jednotliv\u00fdch v\u00fdzv\u00e1ch, na diference ve struktu\u0159e vzorku, na rozd\u00edly ve zp\u016fsobu vypln\u011bn\u00ed dotazn\u00edku (response pattern) a na zm\u011bny v charakteru zji\u0161t\u011bn\u00ed. Hypot\u00e9zy, kter\u00e9 byly v r\u00e1mci jednotliv\u00fdch analytick\u00fdch \u00faloh formulov\u00e1ny, jsou:<\/p>\n\n\n\n

a) V\u00fdb\u011brov\u00fd soubor je reprezentativn\u00ed ve vztahu k z\u00e1kladn\u00ed populaci \u017eadatel\u016f\/p\u0159\u00edjemc\u016f (resp. v\u00fdb\u011brov\u00fd soubor nen\u00ed v\u016f\u010di z\u00e1kladn\u00ed populaci vych\u00fdlen).<\/p>\n\n\n\n

b) Dodate\u010dn\u00e9 v\u00fdzvy vedou ke zv\u00fd\u0161en\u00ed celkov\u00e9 n\u00e1vratnosti.<\/p>\n\n\n\n

c) Dodate\u010dn\u00e9 v\u00fdzvy nemaj\u00ed vliv na strukturu vzorku (resp. rozd\u00edly mezi respondenty, kte\u0159\u00ed reaguj\u00ed na jednotliv\u00e9 v\u00fdzvy, nejsou statisticky v\u00fdznamn\u00e9).<\/p>\n\n\n\n

d) Rozd\u00edly respondent\u016f v p\u0159\u00edstupu k vypl\u0148ov\u00e1n\u00ed dotazn\u00edk\u016f se mezi jednotliv\u00fdmi v\u00fdzvami neli\u0161\u00ed.<\/p>\n\n\n\n

e) Dodate\u010dn\u00e9 v\u00fdzvy nemaj\u00ed vliv na podstatu meritorn\u00edch zji\u0161t\u011bn\u00ed.<\/p>\n\n\n\n

2.2 Metodika<\/strong><\/p>\n\n\n\n

V r\u00e1mci specifikovan\u00e9ho v\u00fdzkumn\u00e9ho \u0161et\u0159en\u00ed byl proveden test \u00fa\u010dinnosti dodate\u010dn\u00fdch v\u00fdzev a anal\u00fdza jejich efekt\u016f, nicm\u00e9n\u011b v\u00fdznamnou ot\u00e1zkou bylo rovn\u011b\u017e ov\u011b\u0159en\u00ed reprezentativity vzorku respondent\u016f. V t\u00e9to souvislosti je t\u0159eba p\u0159edeslat, \u017ee pojem reprezentativity m\u00e1 v relevantn\u00ed literatu\u0159e n\u011bkolik r\u016fzn\u00fdch interpretac\u00ed[12], z nich\u017e n\u011bkter\u00e9 odkazuj\u00ed ke \u201ezmen\u0161enin\u011b populace\u201c \u010di k \u201eabsenci selektivn\u00edch sil\u201c. V kontextu proveden\u00e9ho \u0161et\u0159en\u00ed byla reprezentativita ch\u00e1p\u00e1na jako absence vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f.<\/p>\n\n\n\n

P\u0159i ov\u011b\u0159ov\u00e1n\u00ed reprezentativity se obvykle porovn\u00e1vaj\u00ed v\u00fdzkumn\u00e9 odhady se zn\u00e1m\u00fdmi parametry z\u00e1kladn\u00ed populace, nap\u0159. Thomsen a Siring (1983) takto porovn\u00e1vali \u00fadaje z\u00edskan\u00e9 v r\u00e1mci jimi realizovan\u00e9ho \u0161et\u0159en\u00ed s \u00fadaji ze s\u010d\u00edt\u00e1n\u00ed lidu. V praxi v\u0161ak \u010dasto nast\u00e1v\u00e1 situace, kdy vych\u00fdlen\u00ed nen\u00ed mo\u017en\u00e9 t\u00edmto zp\u016fsobem exaktn\u011b zm\u011b\u0159it vzhledem k tomu, \u017ee nejsou zn\u00e1my skute\u010dn\u00e9 hodnoty parametr\u016f z\u00e1kladn\u00ed populace. Nejinak tomu bylo i v p\u0159\u00edpad\u011b \u0161et\u0159en\u00ed proveden\u00e9ho Institutem evaluac\u00ed a soci\u00e1ln\u00edch anal\u00fdz. V r\u00e1mci konkr\u00e9tn\u00edch okolnost\u00ed sb\u011bru dat v\u0161ak bylo v re\u00e1ln\u00fdch podm\u00ednk\u00e1ch mo\u017en\u00e9 vytvo\u0159it dva na sob\u011b nez\u00e1visl\u00e9 v\u00fdb\u011bry. Vzhledem k tomu, \u017ee n\u00e1vratnost v r\u00e1mci prov\u00e1d\u011bn\u00e9ho \u0161et\u0159en\u00ed[13] dos\u00e1hla ni\u017e\u0161\u00ed ne\u017e o\u010dek\u00e1van\u00e9 m\u00edry, byl ze z\u00e1kladn\u00edho souboru vybr\u00e1n je\u0161t\u011b druh\u00fd v\u00fdb\u011brov\u00fd soubor, jeho\u017e c\u00edlem bylo zv\u00fd\u0161it po\u010det platn\u00fdch p\u0159\u00edpad\u016f pro n\u00e1sledn\u00e9 anal\u00fdzy. T\u00edmto opat\u0159en\u00edm byl jednak napln\u011bn p\u016fvodn\u00ed c\u00edl, tedy zv\u011bt\u0161en\u00ed velikosti zkouman\u00e9ho vzorku, a nav\u00edc tak vnikla mimo\u0159\u00e1dn\u00e1 p\u0159\u00edle\u017eitost k ov\u011b\u0159en\u00ed reprezentativity pomoc\u00ed srovn\u00e1n\u00ed v\u00fdzkumn\u00fdch odhad\u016f zalo\u017een\u00fdch na takto vznikl\u00fdch datov\u00fdch souborech a k verifikaci hypot\u00e9zy o p\u0159\u00edtomnosti, resp. absenci vych\u00fdlen\u00ed.<\/p>\n\n\n\n

Nulov\u00e1 hypot\u00e9za v tomto p\u0159\u00edpad\u011b p\u0159edv\u00edd\u00e1 podobnost v\u00fdzkumn\u00fdch odhad\u016f zalo\u017een\u00fdch na obou v\u00fdb\u011brech, resp. p\u0159edj\u00edm\u00e1, \u017ee p\u0159\u00edpadn\u00e9 rozd\u00edly v jednotliv\u00fdch zji\u0161t\u011bn\u00edch nejsou statisticky v\u00fdznamn\u00e9. Z podobnosti obou v\u00fdb\u011br\u016f lze dovodit, \u017ee by v\u00fdzkumn\u00e9 odhady nevykazovaly statisticky v\u00fdznamn\u00e9 diference ani p\u0159i porovn\u00e1n\u00ed s parametry z\u00e1kladn\u00ed populace (pokud by tyto parametry byly zn\u00e1m\u00e9 a pokud by bylo mo\u017en\u00e9 dan\u00fd test skute\u010dn\u011b prov\u00e9st). Alternativn\u00ed hypot\u00e9za po\u010d\u00edt\u00e1 s existenc\u00ed statisticky v\u00fdznamn\u00fdch odli\u0161nost\u00ed mezi v\u00fdzkumn\u00fdmi odhady zalo\u017een\u00fdmi na jednotliv\u00fdch v\u00fdb\u011brech. Pokud by byly tyto diference prok\u00e1z\u00e1ny, byl by t\u00edm indikov\u00e1n v\u00fdskyt vych\u00fdlen\u00ed. Na z\u00e1klad\u011b dostupn\u00fdch dat by sice nebylo mo\u017en\u00e9 ur\u010dit, kter\u00fd z obou v\u00fdb\u011br\u016f je vych\u00fdlen\u00fd oproti z\u00e1kladn\u00ed populaci, nicm\u00e9n\u011b z hlediska interpretace dosa\u017een\u00fdch v\u00fdsledk\u016f by i takov\u00fdto z\u00e1v\u011br byl u\u017eite\u010dn\u00fd a smyslupln\u00fd.<\/p>\n\n\n\n

V r\u00e1mci obou v\u00fdb\u011brov\u00fdch soubor\u016f byly osloveny v\u0161echny za\u0159azen\u00e9 jednotky. Postup proveden\u00ed sb\u011bru dat byl pro oba v\u00fdb\u011brov\u00e9 soubory identick\u00fd; jedin\u00e1 odli\u0161nost se t\u00fdkala na\u010dasov\u00e1n\u00ed, kdy sb\u011br dat ve v\u00fdb\u011brov\u00fdch souborech neprob\u00edhal paraleln\u011b, n\u00fdbr\u017e sukcesivn\u011b.<\/p>\n\n\n\n

Prvotn\u00ed v\u00fdzva obsahovala \u00favodn\u00ed dopis s vysv\u011btlen\u00edm \u00fa\u010delu a v\u00fdznamu prov\u00e1d\u011bn\u00e9ho \u0161et\u0159en\u00ed, odkaz s mo\u017enost\u00ed ov\u011b\u0159it si v\u011brohodnost prov\u00e1d\u011bn\u00e9ho \u0161et\u0159en\u00ed a apel na zapojen\u00ed do v\u00fdzkumu; incentivy nebyly v tomto p\u0159\u00edpad\u011b nab\u00edzeny.<\/p>\n\n\n\n

3. Zji\u0161t\u011bn\u00e9 v\u00fdsledky<\/strong><\/p>\n\n\n\n

Dosa\u017een\u00e9 v\u00fdsledky jsou prezentov\u00e1ny v p\u011bti \u010d\u00e1stech, kter\u00e9 koresponduj\u00ed s v\u00fd\u0161e formulovan\u00fdmi hypot\u00e9zami. Prvotn\u00ed \u00falohou bylo posoudit reprezentativitu v\u00fdb\u011brov\u00e9ho souboru a odhadnout m\u00edru vych\u00fdlen\u00ed. N\u00e1sledn\u00e9 \u00falohy se zam\u011b\u0159ily na vliv dodate\u010dn\u00fdch v\u00fdzev na celkovou n\u00e1vratnost, na hodnocen\u00ed rozd\u00edl\u016f ve struktu\u0159e v\u00fdsledn\u00e9ho vzorku vyvolan\u00fdch dodate\u010dn\u00fdmi v\u00fdzvami, na posouzen\u00ed rozd\u00edl\u016f ve zp\u016fsobu vypln\u011bn\u00ed on-line dotazn\u00edku mezi respondenty participuj\u00edc\u00edmi na v\u00fdzkumu v r\u00e1mci jednotliv\u00fdch v\u00fdzev a v neposledn\u00ed \u0159ad\u011b tak\u00e9 na anal\u00fdzu hlavn\u00edch zji\u0161t\u011bn\u00ed a zm\u011bn jejich charakteru vyvolan\u00fdch dodate\u010dn\u00fdmi v\u00fdzvami.<\/p>\n\n\n\n

3.1 Ov\u011b\u0159en\u00ed reprezentativity<\/strong><\/p>\n\n\n\n

V r\u00e1mci proveden\u00e9ho \u0161et\u0159en\u00ed bylo zji\u0161\u0165ov\u00e1no celkem 58 prom\u011bnn\u00fdch (bez pomocn\u00fdch a konstruovan\u00fdch znak\u016f) a pro ka\u017edou z nich byly testov\u00e1ny rozd\u00edly v dosa\u017een\u00fdch hodnot\u00e1ch mezi ob\u011bma v\u00fdb\u011brov\u00fdmi soubory. Z hlediska dal\u0161\u00edch anal\u00fdz je t\u0159eba uv\u00e9st, \u017ee toto srovn\u00e1n\u00ed bylo provedeno na z\u00e1klad\u011b dat po obou vln\u00e1ch dodate\u010dn\u00fdch v\u00fdzev. Z\u00e1kladn\u00edm v\u00fdstupem exploratorn\u00ed anal\u00fdzy je histogram, kter\u00fd dokumentuje \u010detnost rozd\u00edl\u016f v hodnot\u00e1ch sledovan\u00fdch prom\u011bnn\u00fdch mezi ob\u011bma v\u00fdb\u011brov\u00fdmi soubory[14]. Je patrn\u00e9, \u017ee hodnoty v\u0161ech sledovan\u00fdch znak\u016f se odchyluj\u00ed od norm\u00e1ln\u00edho rozlo\u017een\u00ed a jsou se\u0161ikmen\u00e9 sm\u011brem k ni\u017e\u0161\u00edm diferen\u010dn\u00edm hodnot\u00e1m. V pr\u016fm\u011bru se jednotliv\u00e9 hodnoty znak\u016f mezi ob\u011bma soubory li\u0161\u00ed o 1,28 procentn\u00edho bodu (st\u0159edn\u00ed hodnota p\u0159itom \u010din\u00ed 0,85 procentn\u00edho bodu). Nejmen\u0161\u00ed pozorovan\u00fd rozd\u00edl \u010din\u00ed 0 procentn\u00edho bodu (jde tedy o identick\u00e9 hodnoty zji\u0161t\u011bn\u00e9 v r\u00e1mci obou v\u00fdb\u011brov\u00fdch soubor\u016f), zat\u00edmco nejv\u011bt\u0161\u00ed diference \u010din\u00ed 4,9 procentn\u00edho bodu. T\u0159i \u010dtvrtiny ze v\u0161ech 146 porovn\u00e1van\u00fdch hodnot se mezi ob\u011bma soubory li\u0161\u00ed nejv\u00fd\u0161e o 1,8 procentn\u00edch bod\u016f.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

N\u00e1sledn\u00e1 anal\u00fdza se zam\u011b\u0159ila na testov\u00e1n\u00ed statistick\u00e9 v\u00fdznamnosti rozd\u00edl\u016f mezi odhady obou v\u00fdb\u011brov\u00fdch soubor\u016f[15]. Z v\u00fdsledk\u016f vyplynulo, \u017ee rozd\u00edly jednotliv\u00fdch znak\u016f mezi ob\u011bma v\u00fdb\u011brov\u00fdmi soubory nejsou v naprost\u00e9 v\u011bt\u0161in\u011b statisticky v\u00fdznamn\u00e9. V\u00fdjimky v tomto ohledu p\u0159edstavuj\u00ed znaky m\u011b\u0159\u00edc\u00ed vn\u00edmanou d\u016fle\u017eitost d\u00edl\u010d\u00edch charakteristik informa\u010dn\u00edho syst\u00e9mu. Z uveden\u00e9ho histogramu (graf 1) je patrn\u00e9, \u017ee diference v hodnot\u00e1ch t\u011bchto znak\u016f p\u0159esahuj\u00ed \u010dty\u0159i procentn\u00ed body. Tyto diference jsou v\u0161ak ojedin\u011bl\u00e9, inkonzistentn\u00ed pokud jde o sm\u011br vych\u00fdlen\u00ed (n\u011bkter\u00e9 hodnoty jsou v druh\u00e9 vln\u011b vy\u0161\u0161\u00ed oproti prvn\u00ed vln\u011b, zat\u00edmco u jin\u00fdch hodnot je tomu p\u0159esn\u011b naopak) a lze je proto pova\u017eovat za nahodil\u00e9 jevy.<\/p>\n\n\n\n

Vzhledem k tomu, \u017ee jde o ekvivalentn\u00ed v\u00fdb\u011bry po\u0159\u00edzen\u00e9 z v\u00fdb\u011brov\u00e9 opory zahrnuj\u00edc\u00ed v\u0161echny jednotky z\u00e1kladn\u00ed populace, je na z\u00e1klad\u011b dosa\u017een\u00fdch v\u00fdsledk\u016f mo\u017en\u00e9 usuzovat, \u017ee oba v\u00fdb\u011brov\u00e9 soubory jsou nevych\u00fdlen\u00e9. Lze rovn\u011b\u017e shrnout, \u017ee z\u00edskan\u00e9 vzorky reprezentuj\u00ed z\u00e1kladn\u00ed populaci a v\u00fdzkumn\u00e9 odhady skute\u010dn\u011b odr\u00e1\u017eej\u00ed parametry z\u00e1kladn\u00ed populace.<\/p>\n\n\n\n

3.2 Zm\u011bna n\u00e1vratnosti po jednotliv\u00fdch v\u00fdzv\u00e1ch<\/strong><\/p>\n\n\n\n

Je patrn\u00e9, \u017ee dodate\u010dn\u00e9 v\u00fdzvy zv\u00fd\u0161ily celkovou n\u00e1vratnost z\u00e1sadn\u00edm zp\u016fsobem, nebo\u0165 v r\u00e1mci prvn\u00edho z v\u00fdb\u011brov\u00fdch soubor\u016f odpov\u011bd\u011blo dokonce v\u00edce respondent\u016f a\u017e na dodate\u010dn\u00e9 v\u00fdzvy (503 respondent\u016f) ne\u017e na \u00favodn\u00ed osloven\u00ed (451 respondent\u016f). Efekt dodate\u010dn\u00fdch v\u00fdzev tedy v p\u0159\u00edpad\u011b prvn\u00edho v\u00fdb\u011brov\u00e9ho souboru \u010din\u00ed 112 procent; n\u00e1vratnost se v tomto p\u0159\u00edpad\u011b toti\u017e zv\u00fd\u0161ila z v\u00fdchoz\u00edch 16,70 % a\u017e na v\u00fdsledn\u00fdch 35,32 %. V r\u00e1mci druh\u00e9ho v\u00fdb\u011brov\u00e9ho souboru \u010din\u00ed celkov\u00fd efekt dodate\u010dn\u00fdch v\u00fdzev 76 %, kdy se n\u00e1vratnost zv\u00fd\u0161ila z v\u00fdchoz\u00edch 18,34 % na v\u00fdsledn\u00fdch 32,32 %. Z tabulky 1 jsou patrn\u00e9 tak\u00e9 absolutn\u00ed po\u010dty vypln\u011bn\u00fdch dotazn\u00edk\u016f. Je tedy patrn\u00e9, \u017ee dodate\u010dn\u00e9 v\u00fdzvy n\u00e1vratnost skute\u010dn\u011b v\u00fdznamn\u011b zvy\u0161uj\u00ed.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Rovn\u011b\u017e je patrn\u00e1 klesaj\u00edc\u00ed mezn\u00ed \u00fa\u010dinnost dodate\u010dn\u00fdch v\u00fdzev,
nebo\u0165 v r\u00e1mci prvn\u00edho v\u00fdb\u011brov\u00e9ho souboru zv\u00fd\u0161ila prvn\u00ed vlna dodate\u010dn\u00fdch v\u00fdzev n\u00e1vratnost o 72 %, zat\u00edmco druh\u00e1 dodate\u010dn\u00e1 v\u00fdzva p\u0159isp\u011bla ke zv\u00fd\u0161en\u00ed o 23 %. V druh\u00e9m v\u00fdb\u011brov\u00e9m souboru zv\u00fd\u0161ila prvn\u00ed v\u00fdzva n\u00e1vratnost o 57 % a druh\u00e1 v\u00fdzva pak n\u00e1vratnost zv\u00fd\u0161ila o dal\u0161\u00edch 12 %.<\/p>\n\n\n\n

Tyto hodnoty jsou vy\u0161\u0161\u00ed, ne\u017e kolik uv\u00e1d\u011bj\u00ed n\u011bkter\u00e9 prameny; nap\u0159. Groves, Cialdini a Couper (1992) pozorovali ve sv\u00e9 studii zv\u00fd\u0161en\u00ed n\u00e1vratnosti v pr\u016fm\u011bru o 11 procent na ka\u017edou vlnu.[16] Ka\u017edop\u00e1dn\u011b je v\u0161ak patrn\u00e9, \u017ee dodate\u010dn\u00e9 v\u00fdzvy maj\u00ed odezvu (by\u0165 postupn\u011b klesaj\u00edc\u00ed) a vedou ke zv\u00fd\u0161en\u00ed celkov\u00e9 n\u00e1vratnosti.<\/p>\n\n\n\n

3.3 Diference ve struktu\u0159e vzorku<\/strong><\/p>\n\n\n\n

V r\u00e1mci verifikace hypot\u00e9zy o vlivu dodate\u010dn\u00fdch v\u00fdzev na reprezentativitu se anal\u00fdza zam\u011b\u0159ila na podrobn\u00fd rozbor v\u00fdzkumn\u00fdch odhad\u016f (hodnot v\u0161ech znak\u016f) v r\u00e1mci obou v\u00fdb\u011brov\u00fdch soubor\u016f. N\u00e1sleduj\u00edc\u00ed anal\u00fdza se obrac\u00ed ke struktu\u0159e soubor\u016f zalo\u017een\u00fdch na odpov\u011bd\u00edch respondent\u016f reaguj\u00edc\u00edch na jednotliv\u00e9 v\u00fdzvy. Pozornost se p\u0159itom soust\u0159e\u010fuje na hlavn\u00ed popisn\u00e9 znaky, jak\u00fdmi jsou typ instituce, v n\u00ed\u017e respondenti pracuj\u00ed, frekvence pr\u00e1ce s informa\u010dn\u00edm syst\u00e9mem BENEFIT7 a v neposledn\u00ed \u0159ad\u011b tak\u00e9 na z\u00e1kladn\u00ed sociodemografick\u00e9 znaky jak\u00fdmi jsou pohlav\u00ed, v\u011bk a nejvy\u0161\u0161\u00ed dosa\u017een\u00e9 vzd\u011bl\u00e1n\u00ed respondent\u016f.<\/p>\n\n\n\n

Z v\u00fdsledk\u016f je patrn\u00e9, \u017ee pracovn\u00edci z ve\u0159ejn\u00e9 spr\u00e1vy reagovali nej\u010dast\u011bji promptn\u011b \u2013 tedy bezprost\u0159edn\u011b na z\u00e1klad\u011b \u00favodn\u00ed pozv\u00e1nky k zapojen\u00ed do v\u00fdzkumu; naproti tomu zam\u011bstnanci v podnikatelsk\u00e9m sektoru reagovali nej\u010dast\u011bji na z\u00e1klad\u011b prvn\u00ed a druh\u00e9 dodate\u010dn\u00e9 v\u00fdzvy. D\u00e1le plat\u00ed, \u017ee pracovn\u00edci nest\u00e1tn\u00edch neziskov\u00fdch organizac\u00ed reagovali typicky a\u017e po druh\u00e9 dodate\u010dn\u00e9 v\u00fdzv\u011b. Rovn\u011b\u017e jsou z\u0159ejm\u00e9 rozd\u00edly dan\u00e9 frekvenc\u00ed pr\u00e1ce s informa\u010dn\u00edm syst\u00e9mem, kdy promptn\u011b odpov\u011bd\u011bli zejm\u00e9na ti, kte\u0159\u00ed se syst\u00e9mem BENEFIT7 pracuj\u00ed \u010dasto (tj. ka\u017ed\u00fd den \u010di n\u011bkolikr\u00e1t t\u00fddn\u011b), zat\u00edmco ostatn\u00ed reagovali a\u017e na prvn\u00ed a druhou dodate\u010dnou v\u00fdzvu. \u017d\u00e1dn\u00e9 dal\u0161\u00ed zm\u011bny ve struktu\u0159e podsoubor\u016f nejsou a\u017e na uveden\u00e9 diference statisticky v\u00fdznamn\u00e9.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

P\u0159esto\u017ee ani dal\u0161\u00ed dva p\u0159\u00edpady diferenc\u00ed nejsou statisticky v\u00fdznamn\u00e9, stoj\u00ed za zm\u00ednku. Uveden\u00e1 zji\u0161t\u011bn\u00ed toti\u017e koresponduj\u00ed se z\u00e1v\u011bry jin\u00fdch \u0161et\u0159en\u00ed (Willke et al. 1999; Lohr 2010), kter\u00e1 dokl\u00e1daj\u00ed rozd\u00edly podle pohlav\u00ed, kdy na dodate\u010dn\u00e9 v\u00fdzvy reaguj\u00ed sp\u00ed\u0161e \u017eeny. Dosa\u017een\u00e9 v\u00fdsledky v tomto ohledu dokl\u00e1daj\u00ed vy\u0161\u0161\u00ed pod\u00edl \u017een reaguj\u00edc\u00edch na druhou dodate\u010dnou v\u00fdzvu (67 %) oproti struktu\u0159e vzorku promptn\u00edch respondent\u016f, v n\u011bm\u017e \u017eeny dosahuj\u00ed pod\u00edlu 61 %. Z dat je d\u00e1le patrn\u00e9, \u017ee star\u0161\u00ed respondenti reaguj\u00ed na prvotn\u00ed v\u00fdzvu ve v\u011bt\u0161\u00ed m\u00ed\u0159e, a tak dodate\u010dn\u00e9 v\u00fdzvy podporuj\u00ed zapojov\u00e1n\u00ed mlad\u0161\u00edch jednotlivc\u016f. Konkr\u00e9tn\u011b se ukazuje, \u017ee zat\u00edmco na \u00favodn\u00ed v\u00fdzvu reagovalo 20 % respondent\u016f star\u0161\u00edch 50 let, na dodate\u010dn\u00e9 v\u00fdzvy se do v\u00fdzkumu nezapojilo v\u00edce ne\u017e 18 % z t\u00e9to v\u011bkov\u00e9 skupiny. V p\u0159\u00edpad\u011b respondent\u016f do 30 let \u010din\u00ed jejich pod\u00edl v r\u00e1mci souboru jednotlivc\u016f reaguj\u00edc\u00edch na prvn\u00ed osloven\u00ed 15 %, zat\u00edmco v druh\u00e9 vln\u011b dodate\u010dn\u00fdch v\u00fdzev jejich pod\u00edl \u010din\u00ed 17 %.<\/p>\n\n\n\n

Z t\u011bchto zji\u0161t\u011bn\u00ed lze tedy dovodit z\u00e1v\u011br, \u017ee dodate\u010dn\u00e9 v\u00fdzvy v konkr\u00e9tn\u00edm situa\u010dn\u00edm kontextu p\u0159isp\u00edvaj\u00ed ke zm\u011bn\u011b struktury vzorku a mohou tedy m\u00edt vliv na vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f. Data prokazuj\u00ed, \u017ee podskupina promptn\u011b reaguj\u00edc\u00edch respondent\u016f se statisticky v\u00fdznamn\u011b odli\u0161uje od t\u011bch, kte\u0159\u00ed reaguj\u00ed a\u017e na z\u00e1klad\u011b dodate\u010dn\u00fdch v\u00fdzev (a to v p\u0159\u00edpad\u011b takov\u00fdch znak\u016f jak\u00fdmi jsou frekvence pr\u00e1ce s informa\u010dn\u00edm syst\u00e9mem BENEFIT7 a typ organizace, v n\u00ed\u017e respondent pracuje). K podobn\u00e9mu z\u00e1v\u011bru dosp\u011bly tak\u00e9 mnoh\u00e9 dal\u0161\u00ed studie, kter\u00e9 prokazuj\u00ed signifikantn\u00ed rozd\u00edl mezi promptn\u00edmi a ostatn\u00edmi respondenty (Dunkelberg, Day 1973; Traugott 1987; Voigt, Koepssell a Daling 2003). Uveden\u00e1 zji\u0161t\u011bn\u00ed tedy ukazuj\u00ed na v\u00fdznam prov\u00e1d\u011bn\u00ed dodate\u010dn\u00fdch v\u00fdzev nejen z hlediska zvy\u0161ov\u00e1n\u00ed n\u00e1vratnosti, ale tak\u00e9 z hlediska zp\u0159es\u0148ov\u00e1n\u00ed v\u00fdzkumn\u00fdch odhad\u016f.<\/p>\n\n\n\n

3.4 Diference ve zp\u016fsobu vypln\u011bn\u00ed dotazn\u00edku<\/strong><\/p>\n\n\n\n

Dal\u0161\u00ed analytick\u00e1 pozornost byla v\u011bnov\u00e1na rozd\u00edl\u016fm ve zp\u016fsobu vypln\u011bn\u00ed on-line dotazn\u00edku (response pattern). V tomto ohledu byly sledov\u00e1ny dv\u011b charakteristiky dokumentuj\u00edc\u00ed p\u0159\u00edstup respondent\u016f k vypln\u011bn\u00ed dotazn\u00edku, a to v\u00fdpadky n\u00e1vratnosti u ot\u00e1zek (item nonresponse), kde bylo mo\u017en\u00e9 se poskytnut\u00ed odpov\u011bdi vyhnout, a celkov\u00fd \u010das v\u011bnovan\u00fd vypl\u0148ov\u00e1n\u00ed dotazn\u00edku.<\/p>\n\n\n\n

a) V\u00fdpadek n\u00e1vratnosti u jednotliv\u00fdch ot\u00e1zek<\/p>\n\n\n\n

P\u0159edm\u011btem srovn\u00e1n\u00ed je po\u010det respondent\u016f, kte\u0159\u00ed neuvedli \u017e\u00e1dnou odpov\u011b\u010f na polo\u017een\u00e9 dotazy. V r\u00e1mci v\u00fdzkumn\u00e9ho instrumentu byly polo\u017eeny celkem t\u0159i samostatn\u00e9 otev\u0159en\u00e9 ot\u00e1zky, v jejich\u017e p\u0159\u00edpad\u011b byla on-line aplikace naprogramov\u00e1na tak, \u017ee je respondent mohl ponechat bez odpov\u011bdi. Po\u010dty respondent\u016f, kte\u0159\u00ed neuvedli \u017e\u00e1dnou odpov\u011b\u010f, jsou patrn\u00e9 z tabulky 3. Plat\u00ed, \u017ee ve skupin\u011b respondent\u016f, kte\u0159\u00ed reagovali bezprost\u0159edn\u011b na \u00favodn\u00ed osloven\u00ed, vyplnilo v\u0161echny otev\u0159en\u00e9 ot\u00e1zky 37,5 % jednotlivc\u016f. Pod\u00edl takto pe\u010dliv\u00fdch respondent\u016f ve skupin\u00e1ch respondent\u016f reaguj\u00edc\u00edch na dodate\u010dn\u00e9 v\u00fdzvy klesl o p\u0159ibli\u017en\u011b \u010dty\u0159i procentn\u00ed body.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

Z konkr\u00e9tn\u00edch dat je tedy z\u0159ejm\u00e1 vy\u0161\u0161\u00ed m\u00edra v\u00fdpadk\u016f n\u00e1vratnosti u dotazn\u00edk\u016f vypln\u011bn\u00fdch v r\u00e1mci dodate\u010dn\u00fdch v\u00fdzev ve srovn\u00e1n\u00ed s dotazn\u00edky vypln\u011bn\u00fdmi promptn\u011b. Bylo by tedy mo\u017en\u00e9 shrnout, \u017ee vzhledem k v\u011bt\u0161\u00edmu po\u010dtu respondent\u016f, kte\u0159\u00ed ponechali otev\u0159en\u00e9 ot\u00e1zky nevypln\u011bn\u00e9, jsou dotazn\u00edky z\u00edskan\u00e9 v r\u00e1mci dodate\u010dn\u00fdch v\u00fdzev jako celek vypln\u011bny m\u00e9n\u011b pe\u010dliv\u011b. Platnost tohoto z\u00e1v\u011bru by m\u011bl dolo\u017eit krat\u0161\u00ed \u010das str\u00e1ven\u00fd vypl\u0148ov\u00e1n\u00edm dotazn\u00edku. Dal\u0161\u00ed anal\u00fdza se proto zam\u011b\u0159ila na porovn\u00e1n\u00ed celkov\u00e9 doby vypl\u0148ov\u00e1n\u00ed.<\/p>\n\n\n\n

b) \u010cas str\u00e1ven\u00fd vypl\u0148ov\u00e1n\u00edm dotazn\u00edku<\/p>\n\n\n\n

V n\u00e1vaznosti na p\u0159edchoz\u00ed z\u00e1v\u011br je d\u016fvodn\u00e9 o\u010dek\u00e1vat, \u017ee \u010das v\u011bnovan\u00fd vypl\u0148ov\u00e1n\u00ed dotazn\u00edku je v r\u00e1mci dodate\u010dn\u00fdch v\u00fdzev krat\u0161\u00ed ne\u017e \u010das, kter\u00fd vypl\u0148ov\u00e1n\u00ed v\u011bnovali promptn\u00ed respondenti. Z dosa\u017een\u00fdch v\u00fdsledk\u016f je v\u0161ak patrn\u00e9, \u017ee se doba vypl\u0148ov\u00e1n\u00ed dotazn\u00edku mezi jednotliv\u00fdmi vlnami li\u0161\u00ed o cca 10 vte\u0159in, co\u017e p\u0159i pr\u016fm\u011brn\u00e9 dob\u011b vypl\u0148ov\u00e1n\u00ed 12 minut nen\u00ed statisticky (a ani v\u011bcn\u011b) v\u00fdznamn\u00fd rozd\u00edl. \u00daspora \u010dasu, ke kter\u00e9 doch\u00e1z\u00ed vynech\u00e1n\u00edm n\u011bkter\u00fdch (otev\u0159en\u00fdch) ot\u00e1zek tedy nevede k celkov\u011b krat\u0161\u00ed dob\u011b vypln\u011bn\u00ed dotazn\u00edku.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Nav\u00edc je na rozd\u00edl od situace u meritorn\u00edch ot\u00e1zek sm\u011brodatn\u00e1 odchylka v r\u00e1mci dodate\u010dn\u00fdch v\u00fdzev vy\u0161\u0161\u00ed ve srovn\u00e1n\u00ed se skupinou promptn\u00edch respondent\u016f (p\u0159ibli\u017en\u011b o jednu minutu), co\u017e ukazuje na v\u011bt\u0161\u00ed rozptyl dat v p\u0159\u00edpad\u011b respondent\u016f zapojuj\u00edc\u00edch se a\u017e na z\u00e1klad\u011b dodate\u010dn\u00fdch v\u00fdzev.<\/p>\n\n\n\n

3.5 Zm\u011bny v charakteru zji\u0161t\u011bn\u00ed<\/strong><\/p>\n\n\n\n

Uk\u00e1zalo se, \u017ee dodate\u010dn\u00e9 v\u00fdzvy nep\u016fsob\u00ed na v\u00fdzkumn\u00e9 odhady jednosm\u011brn\u011b. Zat\u00edmco v n\u011bkter\u00fdch p\u0159\u00edpadech hodnotu z\u00edskanou promptn\u00edmi odpov\u011b\u010fmi sni\u017euj\u00ed, v jin\u00fdch p\u0159\u00edpadech ji naopak zvy\u0161uj\u00ed. Efekt dodate\u010dn\u00fdch v\u00fdzev je nav\u00edc v p\u0159\u00edpad\u011b n\u011bkter\u00fdch prom\u011bnn\u00fdch inkonzistentn\u00ed (resp. neline\u00e1rn\u00ed), nebo\u0165 druh\u00e1 dodate\u010dn\u00e1 v\u00fdzva vychyluje z\u00edskan\u00e9 odhady v opa\u010dn\u00e9m sm\u011bru ne\u017e prvn\u00ed dodate\u010dn\u00e1 v\u00fdzva. V t\u00e9to souvislosti je t\u0159eba uv\u00e9st, \u017ee zji\u0161t\u011bn\u00e9 zm\u011bny nejsou statisticky v\u00fdznamn\u00e9 a nelze tak u hodnocen\u00fdch prom\u011bnn\u00fdch p\u0159i\u010d\u00edst zji\u0161t\u011bn\u00e9 zm\u011bny vlivu dodate\u010dn\u00fdch v\u00fdzev[17]. Popisovan\u00e9 jevy jsou z \u00fasporn\u00fdch d\u016fvod\u016f dokumentov\u00e1ny pouze na souhrnn\u00fdch indik\u00e1torech.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Postupn\u00e9 sni\u017eov\u00e1n\u00ed hodnoty je patrn\u00e9 v p\u0159\u00edpad\u011b indik\u00e1toru spokojenosti s pracovn\u00edm prost\u0159ed\u00edm, kdy dodate\u010dn\u00e9 v\u00fdzvy sn\u00ed\u017eily pr\u016fm\u011brn\u00e9 hodnocen\u00ed z 64,69 % na celkov\u00fdch 64,36 %, a indik\u00e1toru celkov\u00e9 spokojenosti se syst\u00e9mem BENEFIT7, kdy do\u0161lo ke sn\u00ed\u017een\u00ed z 61,58 % na celkov\u00fdch 61,02 %. Inkonzistentn\u00ed vliv v\u00fdzev na v\u00fdzkumn\u00e9 odhady je patrn\u00fd v p\u0159\u00edpad\u011b indik\u00e1tor\u016f spokojenosti s technickou podporou a spokojenosti s daty.<\/p>\n\n\n\n

Z podrobn\u011bj\u0161\u00ed anal\u00fdzy se d\u00e1le ukazuje, \u017ee v r\u00e1mci dodate\u010dn\u00fdch v\u00fdzev (bez ohledu na to, zdali jde o prvn\u00ed \u010di druhou v\u00fdzvu) je sm\u011brodatn\u00e1 odchylka v\u017edy ni\u017e\u0161\u00ed ne\u017e v p\u0159\u00edpad\u011b promptn\u00edch odpov\u011bd\u00ed[18]. Na mo\u017enou p\u0159\u00ed\u010dinu sni\u017euj\u00edc\u00ed se sm\u011brodatn\u00e9 odchylky poukazuje dal\u0161\u00ed z proveden\u00fdch anal\u00fdz, kter\u00e1 dokumentuje postupn\u00e9 zapojov\u00e1n\u00ed n\u00e1zorov\u011b nevyhran\u011bn\u00fdch respondent\u016f. Z dostupn\u00fdch dat lze vysledovat z\u0159etelnou tendenci, kdy n\u00e1zorov\u011b polarizovan\u00ed respondenti (tj. ti, kte\u0159\u00ed jsou velmi spokojeni nebo velmi nespokojeni) reaguj\u00ed promptn\u011b, zat\u00edmco nevyhran\u011bn\u00ed (tedy sp\u00ed\u0161e spokojen\u00ed nebo sp\u00ed\u0161e nespokojen\u00ed) se do v\u00fdzkumu zapojuj\u00ed a\u017e na z\u00e1klad\u011b dodate\u010dn\u00fdch v\u00fdzev.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Z nast\u00edn\u011bn\u00fdch rozd\u00edl\u016f v charakteru zji\u0161t\u011bn\u00ed je patrn\u00e9, \u017ee promptn\u011b reaguj\u00ed sp\u00ed\u0161e respondenti s vyhran\u011bn\u00fdmi n\u00e1zory, resp. lid\u00e9, kte\u0159\u00ed maj\u00ed o dan\u00e9m t\u00e9matu konkr\u00e9tn\u00ed p\u0159edstavu. Naopak jednotlivci s ambivalentn\u00edmi \u010di indiferentn\u00edmi postoji zapojen\u00ed do v\u00fdzkumu odkl\u00e1daj\u00ed a zapojuj\u00ed se a\u017e na z\u00e1klad\u011b dodate\u010dn\u00fdch v\u00fdzev. V\u00fdsledn\u00fd vzorek by tak p\u0159i neproveden\u00ed dodate\u010dn\u00fdch v\u00fdzev nadhodnocoval postojov\u00e9 extr\u00e9my. Podobn\u011b jako v \u010d\u00e1sti v\u011bnovan\u00e9 diferenc\u00edm ve struktu\u0159e vzorku je i zde patrn\u00fd efekt dodate\u010dn\u00fdch v\u00fdzev p\u0159i zp\u0159es\u0148ov\u00e1n\u00ed v\u00fdzkumn\u00fdch odhad\u016f. Na miziv\u00fd rozd\u00edl v \u010dase pot\u0159ebn\u00e9m pro vypln\u011bn\u00ed dotazn\u00edku, kter\u00fd je nast\u00edn\u011bn v\u00fd\u0161e, lze tedy z t\u00e9to perspektivy pohl\u00ed\u017eet jako na indik\u00e1tor vy\u0161\u0161\u00ed kognitivn\u00ed z\u00e1t\u011b\u017ee,
kterou vypln\u011bn\u00ed dotazn\u00edku p\u0159edstavuje pro jedince s nevyhran\u011bn\u00fdmi postoji (resp. s ni\u017e\u0161\u00edm pov\u011bdom\u00edm o p\u0159edm\u011btu v\u00fdzkumu).<\/p>\n\n\n\n

4. Z\u00e1v\u011bry<\/strong><\/p>\n\n\n\n

Na z\u00e1klad\u011b dosa\u017een\u00fdch v\u00fdsledk\u016f je tedy mo\u017en\u00e9 konstatovat, \u017ee n\u00edzk\u00e1 m\u00edra n\u00e1vratnosti nemus\u00ed sama o sob\u011b nutn\u011b znamenat, \u017ee jsou v\u00fdzkumn\u00e9 odhady vych\u00fdlen\u00e9 a zji\u0161t\u011bn\u00e9 \u00fadaje nereprezentativn\u00ed. Z v\u00fdsledk\u016f mnou prov\u00e1d\u011bn\u00fdch anal\u00fdz je patrn\u00e9, \u017ee vzorek promptn\u00edch respondent\u016f (tj. v r\u00e1mci proveden\u00e9ho \u0161et\u0159en\u00ed n\u00e1vratnost 17\u201318 %) dok\u00e1\u017ee reprezentovat z\u00e1kladn\u00ed populaci stejn\u011b dob\u0159e jako vzorek dopln\u011bn\u00fd o respondenty reaguj\u00edc\u00ed na dodate\u010dn\u00e9 v\u00fdzvy. I kdy\u017e n\u011bkter\u00e9 prameny poukazuj\u00ed na p\u0159ijatelnou m\u00edru n\u00e1vratnosti (mj. Babbie 2007), nelze shrnout \u010di uv\u00e9st,
jak\u00e1 m\u00edra n\u00e1vratnosti je adekv\u00e1tn\u00ed \u010di dostate\u010dn\u00e1, nebo\u0165 z\u00e1le\u017e\u00ed mj. na povaze pasivn\u00edch jednotlivc\u016f. Pokud je nap\u0159. skupina pasivn\u00edch jednotlivc\u016f velmi heterogenn\u00ed, tak i 90% n\u00e1vratnost m\u016f\u017ee b\u00fdt nedostate\u010dn\u00e1, a naopak v mnoh\u00fdch jin\u00fdch p\u0159\u00edpadech nemus\u00ed 20% n\u00e1vratnost p\u0159edstavovat sebemen\u0161\u00ed probl\u00e9m. P\u0159esto je \u00fadaj o n\u00e1vratnosti d\u016fle\u017eitou informac\u00ed, kter\u00e1 by m\u011bla b\u00fdt uv\u00e1d\u011bna spolu s dal\u0161\u00edmi informacemi p\u0159i prezentaci v\u00fdsledk\u016f z prov\u00e1d\u011bn\u00fdch \u0161et\u0159en\u00ed.<\/p>\n\n\n\n

Dosa\u017een\u00e9 v\u00fdsledky rovn\u011b\u017e potvrzuj\u00ed zn\u00e1m\u00fd fakt, \u017ee velikost zkouman\u00e9ho souboru nem\u00e1 na reprezentativitu p\u0159\u00edm\u00fd vliv. Nicm\u00e9n\u011b vy\u0161\u0161\u00ed po\u010det jednotek ve vzorku poskytuje lep\u0161\u00ed analytick\u00e9 mo\u017enosti, nebo\u0165 s vy\u0161\u0161\u00edm po\u010dtem p\u0159\u00edpad\u016f lze prov\u00e1d\u011bt detailn\u011bj\u0161\u00ed anal\u00fdzy \u010di konstruovat p\u0159esn\u011bji vymezen\u00e9 segmenty. Konkr\u00e9tn\u00ed m\u00edru n\u00e1vratnosti je tedy vhodn\u00e9 posuzovat v kontextu homogenity \u010di heterogenity z\u00e1kladn\u00ed populace; pokud je zkouman\u00e1 populace fragmentov\u00e1na, je pou\u017eit\u00ed n\u00e1stroj\u016f pro zv\u00fd\u0161en\u00ed n\u00e1vratnosti vhodn\u00e9.<\/p>\n\n\n\n

Dal\u0161\u00edm v\u00fdznamn\u00fdm z\u00e1v\u011brem je, \u017ee dodate\u010dn\u00e9 v\u00fdzvy nezp\u016fsobily v podm\u00ednk\u00e1ch proveden\u00e9ho \u0161et\u0159en\u00ed vych\u00fdlen\u00ed dat, resp. \u017ee zji\u0161t\u011bn\u00e9 odchylky jsou nahodil\u00e9 a nemaj\u00ed charakter systematick\u00e9ho vych\u00fdlen\u00ed. Nicm\u00e9n\u011b v d\u00edl\u010d\u00edch parametrech je patrn\u00e1 struktur\u00e1ln\u00ed distorze v\u00fdb\u011brov\u00e9ho souboru (v r\u00e1mci proveden\u00e9ho \u0161et\u0159en\u00ed dle v\u011bku a dle frekvence pr\u00e1ce s informa\u010dn\u00edm syst\u00e9mem). V tomto ohledu jsou dosa\u017een\u00e9 v\u00fdsledky podobn\u00e9 s v\u00fdsledky nap\u0159. Willkeho et al. (1999), kte\u0159\u00ed uv\u00e1d\u011bj\u00ed, \u017ee v jejich \u0161et\u0159en\u00ed byla struktura respondent\u016f reaguj\u00edc\u00edch na dodate\u010dn\u00e9 v\u00fdzvy jin\u00e1 ve srovn\u00e1n\u00ed se skupinou reaguj\u00edc\u00ed na \u00favodn\u00ed osloven\u00ed; k podobn\u00e9mu z\u00e1v\u011bru dosp\u011bl mj. tak\u00e9 Lohr (2010).<\/p>\n\n\n\n

Nav\u00edc plat\u00ed, \u017ee efekt v\u00fdzev nemus\u00ed b\u00fdt jen pozitivn\u00ed (jednosm\u011brn\u00fd); jin\u00fdmi slovy nemus\u00ed platit p\u0159\u00edm\u00e1 \u00fam\u011bra, \u017ee \u010d\u00edm v\u00edce v\u00fdzev, t\u00edm lep\u0161\u00ed data. Dodate\u010dn\u00e9 v\u00fdzvy toti\u017e mohou strukturu v\u00fdb\u011brov\u00e9ho souboru vych\u00fdlit,
nebo\u0165 reaktance na opakovan\u00e9 v\u00fdzvy se li\u0161\u00ed podle typu respondenta. Z proveden\u00e9ho v\u00fdzkumu se mj. ukazuje, \u017ee na opakovan\u00e9 v\u00fdzvy reaguj\u00ed ve v\u011bt\u0161\u00ed m\u00ed\u0159e \u017eeny; podobn\u011b tak\u00e9 n\u00e1zorov\u011b nevyhran\u011bn\u00ed (ambivalentn\u00ed) respondenti se zapojuj\u00ed zejm\u00e9na a\u017e na z\u00e1klad\u011b dodate\u010dn\u00fdch v\u00fdzev. Dodate\u010dn\u00e9 v\u00fdzvy tedy mohou ovlivnit rozptyl, a pota\u017emo i p\u0159esnost v\u00fdsledk\u016f. P\u0159i jejich pou\u017eit\u00ed je tedy v\u0159ele doporu\u010deno d\u016fkladn\u011b analyzovat odli\u0161nosti v\u00fdzkumn\u00fdch odhad\u016f mezi jednotliv\u00fdmi vlnami dodate\u010dn\u00fdch v\u00fdzev.<\/p>\n\n\n\n

Po\u010det dvou dodate\u010dn\u00fdch v\u00fdzev n\u00e1sleduj\u00edc\u00edch po \u00favodn\u00edm osloven\u00ed se uk\u00e1zal optim\u00e1ln\u00edm kompromisem mezi snahou o maximalizaci n\u00e1vratnosti a ne\u00fanosnou obtrusivitou z pohledu respondent\u016f. Je t\u0159eba p\u0159ipomenout, \u017ee dle platn\u00fdch etick\u00fdch kodex\u016f upravuj\u00edc\u00edch prov\u00e1d\u011bn\u00ed v\u00fdzkumn\u00fdch \u0161et\u0159en\u00ed (AAPOR 2015) je zapojen\u00ed do v\u00fdzkumu dobrovoln\u00e9 a nelze ho vynucovat. Prvotn\u00ed osloven\u00ed n\u00e1sledovan\u00e9 dv\u011bma opakovan\u00fdmi v\u00fdzvami dobrovolnost respondent\u016f participovat na v\u00fdzkumu respektuje a p\u0159edstavuje pro osloven\u00e9 jednotlivce p\u0159ijateln\u00fd \u201en\u00e1tlak\u201c. V p\u0159\u00edpad\u011b CAWI nejsou n\u00e1klady na dodate\u010dn\u00e9 v\u00fdzvy podstatn\u00fdm v\u00fddajem, nicm\u00e9n\u011b v\u00edce ne\u017e dv\u011b vlny lze v r\u00e1mci dan\u00e9 c\u00edlov\u00e9 skupiny pova\u017eovat za kontraproduktivn\u00ed. Jak je patrn\u00e9 z konkr\u00e9tn\u00edch v\u00fdsledk\u016f proveden\u00e9ho \u0161et\u0159en\u00ed, nezvy\u0161uje se m\u00edra n\u00e1vratnosti konstantn\u011b. Na z\u00e1klad\u011b porovn\u00e1n\u00ed \u00fa\u010dinnosti jednotliv\u00fdch dodate\u010dn\u00fdch v\u00fdzev lze odhadnout, \u017ee t\u0159et\u00ed vlna by patrn\u011b nep\u0159inesla v\u00edce ne\u017e 1\u20133 % odpov\u011bd\u00ed (cca 50 respondent\u016f), co\u017e je zcela margin\u00e1ln\u00ed efekt. Lze doporu\u010dit, aby problematice optim\u00e1ln\u00edho po\u010dtu dodate\u010dn\u00fdch v\u00fdzev a prok\u00e1z\u00e1n\u00ed jejich \u00fa\u010dink\u016f byla v\u011bnov\u00e1na dal\u0161\u00ed pozornost evalu\u00e1tor\u016f v r\u00e1mci p\u0159ipravovan\u00fdch \u0161et\u0159en\u00ed.<\/p>\n\n\n\n

Zdroje<\/strong><\/p>\n\n\n\n

[1]        Aapor (2015) Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. 8th edition. AAPOR<\/p>\n\n\n\n

[2]        Atrostic, B. K., Bates, N., Burt, G., Silberstein, A., Winters, F. (1999) \u201cNonresponse in Household Surveys: New Measures and Insights\u201d, a paper presented at the International Conference on Survey Nonresponse, Portland, Oregon.<\/p>\n\n\n\n

[3]        Babbie, E. R. (2007) The Practice of Social Research. Belmont: Wadsworth.<\/p>\n\n\n\n

[4]        Bailar, B. A., Lanphier, C. M. (1978) Development of Survey Methods to Assess Survey Practices. Washington, D. C.: American Statistical Association.<\/p>\n\n\n\n

[5]        Batagelj, Z., Lozar, K., Vehovar, V. (1998) Respondent\u00b4s Satisfaction in WWW Surveys. Paper presented at the International Conference on Methodology and Statistics, Preddvor, Slovenia.<\/p>\n\n\n\n

[6]        Bates, N., Creighton, K. (2000) The Last Five Percent: What Can We Learn from Difficult Interviews? In: Proceedings of the Annual Meetings of the American Statistical Association, 13\u201317 August.<\/p>\n\n\n\n

[7]        Biemer, P., Lyberg, L. (2003) Introduction to Survey Quality. Wiley Series in Survey Methodology.<\/p>\n\n\n\n

[8]        Brick, J. M., Bose, J. (2001) Analysis of potential nonresponse bias. Proceedings of the Joint Statistical Meetings, American Statistical Association, Atlanta.<\/p>\n\n\n\n

[9]        Brownlee, K. A. (1957) A Note of the Effects of Nonresponse on Surveys. American Statistical Association Journal, 29\u201332.<\/p>\n\n\n\n

[10]      Curtin, R., Presser, S., Singer, E. (2005) Changes in Telephone Survey Nonresponse over the Past Quarter Century. Public Opinion Quarterly, Volum 69, pp. 87\u201398.<\/p>\n\n\n\n

[11]      de Leeuw, E., de Heer, W. (2002) Trends in Household Survey Nonresponse: A Longitudinal and International Perspective. In: Survey Nonresponse. New York: Wiley, pp. 41\u201354.<\/p>\n\n\n\n

[12]      Deming, W. E. (1953) On the Probability Mechanism to Attain an Economic Balance Between the Resultant Error of Non-Response and the Bias of Non-Response. Journal of the American Statistical Association, 48, 743\u2013772.<\/p>\n\n\n\n

[13]      Dillman, D. A., (1991) The Design and Administration of Mail Surveys. In: Annual Review of Sociology. Vol. 17, pp. 225\u2013249.<\/p>\n\n\n\n

[14]      Dillman, D. A., (1972) Increasing Mail Questionnaire Response in Large Samples of the General Public. Public Opinion Quarterly. Vol. 36, pp. 257\u2013257.<\/p>\n\n\n\n

[15]      Dohremwend, B. S. (1970) An Experimental Study of Payments to Respondents. Public Opinion Quarterly. Vol. 34, pp. 621\u2013624.<\/p>\n\n\n\n

[16]      Donald, M. N. (1960) Implications of Nonresponse for the Interpretation of Mail Questionnaire Data. Public Opinion Quarterly. Vol. 24, pp. 99\u2013114.<\/p>\n\n\n\n

[17]      Dunkelberg,W. C., and Day, G. S. (1973) Nonresponse Bias and Callbacks in Sample Surveys In: Journal of Marketing Research. Vol 10, pp. 160\u2013168.<\/p>\n\n\n\n

[18]      Enander, J., Sajti, A. (1999) Online Survey of Online Customers, Value-Added Market Research through Data Collection on Internet. In: Proceedings of the ESOMAR Worldwide Internet Conference Net Effects. Amsterdam: ESOMAR, pp. 35\u201352.<\/p>\n\n\n\n

[19]      Ferber, R., Sudman S. (1974) Effects of Compensation in Consumer Expenditure Surveys. Annals of Economic and Social Measurement. Vol. 3, pp. 319\u2013331.<\/p>\n\n\n\n

[20]      Flemming, G., Sonner, M. (1999) Can Internet Polling Work? Strategies for Conducting Public Opinion Surveys Online, paper presented at the 1999 AAPOR Conference, St. Petersburg, Florida.<\/p>\n\n\n\n

[21]      Groves, R. M., Fowler, F. J., Couper, M. P. (2004) Survey Methodology. Hoboken: John Wiley and Sons, Inc.<\/p>\n\n\n\n

[22]      Groves, R. M., Couper, M. (1998) Nonresponse in Household Interview Surveys. John Wiley and Sons, Inc.: New York, NY.<\/p>\n\n\n\n

[23]      Groves, R. M., Dillman, D. A., Eltinge, J. L., Little, R. A. (2002) Survey Nonresponse. New York: John Wiley and Sons, Inc.<\/p>\n\n\n\n

[24]      Groves, R., Cialdini, R., Couper, M. (1992) \u201cUnderstanding the Decision to Participate in a Survey\u201d, Public Opinion Quarterly, Vol. 56, No. 4, Winter, pp. 475 -95.<\/p>\n\n\n\n

[25]      Groves, R., Peytcheva, P. (2008) \u201cThe Impact of Nonresponse Rates on Nonresponse Bias: A Meta-Analysis.\u201d Public Opinion Quarterly 72(2):167\u201389.<\/p>\n\n\n\n

[26]      Groves, R., Singer, E., Corning, A. (2000) Leverage-saliency Theory of Survey Participation: Description and an Illustration. Public Opinion Quarterly, Vol. 64, pp. 299\u2013308.<\/p>\n\n\n\n

[27]      Gunn, W., Rhodes, I. (1981) Physician Response Rates to a Telephone Survey: Effects of Monetary Incetive Level. In: Public Opinion Quarterly, Vol. 45, pp. 109\u2013115.<\/p>\n\n\n\n

[28]      J\u00e4rnbert, M., \u00d6hrvall, R. (2012) Effects of Nonresponse on Survey Estimates of Political Participation. Paper presented at the European Conference on Quality in Official Statistics. Athens, Greece<\/p>\n\n\n\n

[29]      Keeter, S., Miller, C., Kohut, A., Groves, R. M. and Presser, S. (2000) Consequences of reducing nonresponse in a national telephone survey. Public Opinion Quarterly. Vol. 64, pp. 125\u2013148.<\/p>\n\n\n\n

[30]      Kruskal, W., Mosteller, F. (1979) Representative sampling III: Current statistical literature. International Statistical Review, 47, 245-265.<\/p>\n\n\n\n

[31]      Krej\u010d\u00ed, J. (2008) Kvalita soci\u00e1ln\u011bv\u011bdn\u00edch v\u00fdb\u011brov\u00fdch \u0161et\u0159en\u00ed v \u010cesk\u00e9 republice. Praha: Sociologick\u00e9 Nakladatelstv\u00ed.<\/p>\n\n\n\n

[32]      Lohr, S. L. (2010) Sampling:Design and Analysis. Boston: Brooks\/Cole.<\/p>\n\n\n\n

[33]      Lessler, J. T., Kalsbeek, W. D. (1992) Nonsampling Error in Surveys. John Wiley and Sons, Inc.: New York, NY.<\/p>\n\n\n\n

[34]      Madow, W. G., Nisselson, H. and Olkin, I. (eds) (1983) Incomplete Data in Sample Surveys. Vol 3. New York Academic Press.<\/p>\n\n\n\n

[35]      Mehta, R., Sivadas, E. (1995) Comparing Response Rates and Response Content in Mail Versus Electronic Mail Surveys. In: Journal of the Market Research Society, Vol. 37, pp. 429\u2013439.<\/p>\n\n\n\n

[36]      Merkle, D. M., and Edelman, M. (2002) Nonresponse in exit polls: A comprehensive analysis. In Survey Nonresponse (Eds. R. M. Groves, D. A. Dillman, J. L. Eltinge and R. J. A Little). New York: John Wiley & Sons, Inc., 243\u2013258.<\/p>\n\n\n\n

[37]      S\u00e4rndal, C. E., Lundstr\u00f6m, S. (2005) Estimation in Surveys with Nonresponse. Chichester: Wiley.<\/p>\n\n\n\n

[38]      Sheehan, K. B., Hoy, M. G. (1999) Using E-mail to Survey Internet Users in the United States: Methodology and Assessment. In: Journal of Computer Mediated Communication, Vol 4, p. 3.<\/p>\n\n\n\n

[39]      Schouten, B. (2004) Adjustment for bias in the Integrated Survey on Household Living Conditions (POLS) 1998. Discussion paper 04001, CBS, Voorburg<\/p>\n\n\n\n

[40]      Steeh, C. (1981) Trends in Response Rates. Public Opinion Quarterly, 45, 40\u201357.<\/p>\n\n\n\n

[41]      Thomsen, I. and Siring, E. (1983) On the Causes and Effects of Nonresponse: Norwegian Experiences In: W. G. Madow and I Olkin (eds), Incomplete Data in Sample Surveys, Vol. 3. New York; Academic Press.<\/p>\n\n\n\n

[42]      Thomsen, I., Kleven, O., Wang, J. H., Zhang, L. C. (2006) Coping with Decreasing Nonresponse Rates in Statistics Norway. Oslo, Statistics Norway.<\/p>\n\n\n\n

[43]      Traugott, M. W. (1987) The Importance of Persistence in Respondent Selection for Preelection Surveys In: Public Opinion Quarterly, 51, 48\u201357.<\/p>\n\n\n\n

[44]      Voigt, L. F., Koepssell, T. D. and Daling, J. R. (2003) Characteristics of Telephone Survey Respondents According toWillingness to Participate In: American Journal of Epidemiology, 157, 66\u201373.<\/p>\n\n\n\n

[45]      Willke, J., Adams, C., Girnius, Z. (1999) Internet Testing. A Landmark Study of the Differences Between Mail Intercept and On-line Interviewing in the United States. In: Proceeding of the Worldwide Internet Seminar 1998 in Paris, France, Amsterdam: ESOMAR, pp. 145\u2013157.<\/p>\n\n\n\n

\n\n\n\n



[1] Anglick\u00fd pojem non-response nem\u00e1 adekv\u00e1tn\u00ed \u010desk\u00fd ekvivalent, a proto v dal\u0161\u00edm textu pou\u017e\u00edv\u00e1m pojem v\u00fdpadek n\u00e1vratnosti navrhovan\u00fd Krej\u010d\u00edm (2008).
[2] Markantn\u00ed je srovn\u00e1n\u00ed typick\u00e9 n\u00e1vratnosti mezi \u0161et\u0159en\u00edmi prov\u00e1d\u011bn\u00fdmi metodou CAWI (Computer Assisted Web Interviewing, tedy pomoc\u00ed on-line dotazn\u00edk\u016f, kter\u00e9 vypl\u0148uj\u00ed jednotliv\u00ed respondenti samostatn\u011b) a face-to-face interview. Zvy\u0161uj\u00edc\u00ed se pod\u00edl v\u00fdzkum\u016f prov\u00e1d\u011bn\u00fdch metodou CAWI, pro kter\u00e9 je ni\u017e\u0161\u00ed n\u00e1vratnost p\u0159irozen\u00e1, tak m\u016f\u017ee p\u0159isp\u011bt ke klesaj\u00edc\u00edmu trendu vykazovan\u00e9 n\u00e1vratnosti.
[3] To mnohdy nemus\u00ed korespondovat s v\u00fd\u010dtem jednotek ve v\u00fdb\u011brov\u00e9 opo\u0159e, nebo\u0165 n\u011bkter\u00e9 v\u00fdb\u011brov\u00e9 opory zahrnuj\u00ed i jednotky, kter\u00e9 jsou za\u0159azeny omylem nebo neodpov\u00eddaj\u00ed vymezen\u00ed c\u00edlov\u00e9 skupiny (co\u017e se zjist\u00ed a\u017e b\u011bhem samotn\u00e9ho dotazov\u00e1n\u00ed, nap\u0159. b\u011bhem screeningu).
[4] AAPOR (American Association for Public Opinion Research) definuje celkem \u0161est m\u011br n\u00e1vratnosti; v tomto textu je n\u00e1vratnost kalkulov\u00e1na dle standardu RR2.
[5] V d\u016fsledku by tento pocit mohl ovlivnit nejen n\u00e1vratnost, ale i vych\u00fdlen\u00ed v\u00fdzkumn\u00fdch odhad\u016f.
[6] Jde zejm\u00e9na o techniku v\u00fdb\u011bru, zvolenou metodu sb\u011bru dat a \u010das vymezen\u00fd na ter\u00e9nn\u00ed pr\u00e1ce.
[7] CATI \u2013 Computer Assisted Telephone Interviewing
[8] Vych\u00fdlen\u00ed (bias) je rozd\u00edl mezi v\u00fdzkumn\u00fdm odhadem a parametrem z\u00e1kladn\u00ed populace, resp. skute\u010dnou hodnotou (Brick, Bose 2001).
[9] V\u00edce o mo\u017enostech korekce zji\u0161t\u011bn\u00ed pojedn\u00e1v\u00e1 nap\u0159. Lessler a Kalsbeek (1992).
[10] Jde o specifick\u00fd typ uspo\u0159\u00e1d\u00e1n\u00ed sb\u011bru dat, kter\u00fd poskytuje respondentovi mo\u017enost volby mezi n\u011bkolika metodami sb\u011bru dat.
[11] Aplikace dodate\u010dn\u00fdch v\u00fdzev v\u0161ak nen\u00ed v\u00e1z\u00e1na v\u00fdlu\u010dn\u011b na CAWI; stejn\u011b dob\u0159e mohou b\u00fdt dodate\u010dn\u00e9 v\u00fdzvy pou\u017eity i p\u0159i jin\u00fdch metod\u00e1ch sb\u011bru dat.
[12] Jejich p\u0159ehled viz Kruskal a Mosteller (1979).
[13] Po\u010det osloven\u00fdch respondent\u016f byl z\u00e1m\u011brn\u011b relativn\u011b n\u00edzk\u00fd vzhledem ke snaze minimalizovat z\u00e1t\u011b\u017e zkouman\u00e9 populace \u2013 v p\u0159\u00edpad\u011b \u017eadatel\u016f\/p\u0159\u00edjemc\u016f pou\u017e\u00edvaj\u00edc\u00edch informa\u010dn\u00ed syst\u00e9m BENEFIT7 je toti\u017e riziko v \u00favodu zm\u00edn\u011bn\u00e9 p\u0159ezkoumanosti pom\u011brn\u011b vysok\u00e9.
[14] N je v tomto p\u0159\u00edpad\u011b rovno celkov\u00e9mu po\u010dtu porovn\u00e1van\u00fdch hodnot; celkem bylo porovn\u00e1no 146 hodnot.
[15] Testov\u00e1ny byly diference mezi v\u0161emi 58 prom\u011bnn\u00fdmi; rozd\u00edly mezi nomin\u00e1ln\u00edmi znaky byly testov\u00e1ny pomoc\u00ed ch\u00ed-kvadr\u00e1tov\u00e9ho testu, v p\u0159\u00edpad\u011b ordin\u00e1ln\u00edch a kardin\u00e1ln\u00edch prom\u011bnn\u00fdch byl pou\u017eit Mann\u2013Whitney\u016fv U-test.
[16] Podobn\u011b tak\u00e9 Enander, Sajti (1999); Batagelj, Lozar a Vehovar (1998) \u010di Flemming, Sonner (1999) ve sv\u00fdch v\u00fdsledc\u00edch uv\u00e1d\u011bj\u00ed, \u017ee dodate\u010dn\u00e9 v\u00fdzvy zv\u00fd\u0161ily v jimi prov\u00e1d\u011bn\u00fdch \u0161et\u0159en\u00edch n\u00e1vratnost p\u0159ibli\u017en\u011b o t\u0159etinu.
[17] Jde tedy o nev\u00fdznamn\u00fd jev, jeho\u017e efekt by sice bylo mo\u017en\u00e9 predikovat pomoc\u00ed regresn\u00edho modelu, nicm\u00e9n\u011b praktick\u00fd smysl by takov\u00e1to anal\u00fdza nem\u011bla, nebo\u0165 jde o unik\u00e1tn\u00ed, nezobecniteln\u00fd vliv t\u00fdkaj\u00edc\u00ed se konkr\u00e9tn\u00edch znak\u016f.
[18] V t\u00e9to souvislosti je t\u0159eba upozornit, \u017ee ka\u017ed\u00e1 vlna byla v\u017edy hodnocena samostatn\u011b. Sni\u017eov\u00e1n\u00ed sm\u011brodatn\u00e9 odchylky (a pota\u017emo i rozptylu) tedy nem\u016f\u017ee b\u00fdt zp\u016fsobeno vy\u0161\u0161\u00edm po\u010dtem p\u0159\u00edpad\u016f.<\/p>\n","protected":false},"excerpt":{"rendered":"

Text turns the attention to nonresponse that is the one of the key characteristic of both evaluation and research inquiries. Besides the review of most important definitions of nonresponse, an overview of selected approaches for improvement the response rate is presented. Moreover, the possible changes to research design are introduced and discussed. However, the main attention is paid to follow-ups, their usefulness and impact on response rate and bias of the research estimates. Text also demonstrates how the follow-ups can influence the structure of the research sample and how it affects the character of achieved results. Specific attention is paid to answering the question of optimal number of follow-up waves. Presented conclusions are based on empirical data describing the satisfaction with information system BENFIT7; such data were collected by Institute for evaluations and social analyses with the use of CAWI.<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[26],"tags":[],"class_list":["post-1406","post","type-post","status-publish","format-standard","hentry","category-articles"],"acf":[],"_links":{"self":[{"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/posts\/1406","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/comments?post=1406"}],"version-history":[{"count":2,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/posts\/1406\/revisions"}],"predecessor-version":[{"id":1408,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/posts\/1406\/revisions\/1408"}],"wp:attachment":[{"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/media?parent=1406"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/categories?post=1406"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/evaltep.xcreative.cz\/en\/wp-json\/wp\/v2\/tags?post=1406"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}