Multivarijaciona analiza višesmernih tabela kontingencije u funkciji uzročne analize
Analyzing causes by multivariate analysis of multi-way tables of contingency
dc.creator | Cvejić, Slobodan | |
dc.date.accessioned | 2021-10-12T10:17:46Z | |
dc.date.available | 2021-10-12T10:17:46Z | |
dc.date.issued | 1997 | |
dc.identifier.issn | 0085-6320 | |
dc.identifier.uri | http://reff.f.bg.ac.rs/handle/123456789/247 | |
dc.description.abstract | Loglinearnim pristupom modeliramo ćelijske frekvencije u tabeli kontigencije na osnovu pretpostavljenih veza između promenljivih. Na ovaj način se iz kompleksa međuveza nastoje odstraniti prividne veze da bi se došlo do značajnih i smislenih veza između pojava. Loglinearni modeli ne mogu pomoći istraživačima da dosegnu imaginaciju kakvu je imao Durkheim, ali mogu izvrsno instrumentalizovati njihove analitičke sposobnosti i olakšati im prvi korak od istraživanja ka teoriji. Prednosti upotrebe loglinearnih modela nisu potvrđene samo njihovom širokom upotrebom, nego i konsenzusom brojnih metodologa. Logit analiza služi za ispitivanje relacije između dihotomne zavisne i jedne ili više nezavisnih varijabli. I ona je bazirana na proučavanju tabela kontigencije, ali ovde zavisnu promenljivu ne predstavljaju očekivane frekvencije nego logaritmovane šanse (log odds) za koje se koristi naziv logit. Još veći značaj i širi domen primene logit analize proističe iz njene primenljivosti na slučajeve kada zavisna varijabla nije dihotomna (k gt 2). Tada se konstruiše niz logit modela čije se performanse porede. Ovakvim pristupom je moguće analizirati ne samo varijable, nego i njihove kategorije, što nipošto nije od malog značaja u sociološkim istraživanjima. | sr |
dc.description.abstract | Using loglinear approach, we model frequencies in contingency table according to assumed relations between variables. This way we remove false relations from set of interrelations to find out which relations are true and significant. Loglinear models can't help researchers achieve imagination Durkheim had, but they can instrumentalise their analytical abilities and ease their first step from research towards theory. The advantage of use of loglinear models are approved not only by their every day use, but also by many methodologists consensus.Logit analysis has been used to investigate relations between dichotomous dependent variable and one or more independent variables. It is based on contingency tables research too, but here dependent variable is not represented by expected frequency, but by log odds called logit. Even more importance and wider range of use of logit models comes out from their applicability to situations where dependent variable is not dichotomous (k gt 2). That's when series of logit models are constructed and compared. This approach allows analysis of not only variables, but their categories too, which has great importance in sociological researches. | en |
dc.publisher | Sociološko društvo Srbije, Beograd | |
dc.rights | openAccess | |
dc.rights.uri | https://creativecommons.org/licenses/by-sa/4.0/ | |
dc.source | Sociološki pregled | |
dc.subject | varijable | sr |
dc.subject | uzročnost | sr |
dc.subject | tabele kontingencije | sr |
dc.subject | prilagođenost modela | sr |
dc.subject | povezanost | sr |
dc.subject | frekvencije | sr |
dc.subject | variables | en |
dc.subject | tables of contingency | en |
dc.subject | frequencies | en |
dc.subject | connections | en |
dc.subject | causality | en |
dc.subject | adjustability of models | en |
dc.title | Multivarijaciona analiza višesmernih tabela kontingencije u funkciji uzročne analize | sr |
dc.title | Analyzing causes by multivariate analysis of multi-way tables of contingency | en |
dc.type | article | |
dc.rights.license | BY-SA | |
dc.citation.epage | 496 | |
dc.citation.issue | 4 | |
dc.citation.other | 31(4): 477-496 | |
dc.citation.spage | 477 | |
dc.citation.volume | 31 | |
dc.identifier.fulltext | http://reff.f.bg.ac.rs/bitstream/id/16616/bitstream_16616.pdf | |
dc.identifier.rcub | https://hdl.handle.net/21.15107/rcub_reff_247 | |
dc.type.version | publishedVersion |