摘要:An experiment in the field of social psychology is presented,emphasising the application of some less renowned statistical methods for analysing ranked data.The experiment addresses two principal issues: the association between norm congruency and formation of peer groups and the internal (personal) vs.the external (behavioural) aspect of norms.This calls for analysing agreement within and between several groups of judges ranking several objects.After an overview of the interrelations between different forms and measures of rank-correlation and concordance is provided,the problem is tackled by means of ANACONDA (a procedure analogous to ANOVA,developed by Schucany and associates in the nineteen-seventies),whereby a solution is proposed for the previously unaddressed issue of testing overall (factor) effects.Finally,estimation of confidence intervals for concordance coefficient in the non-null case is illustrated.
其他摘要:V članku je predstavljen eksperiment s področja socialne psihologije,pri čemer je izpostavljena uporaba manj znanih statističnih metod za analizo podatkov v obliki rangov.Eksperiment je namenjen preučevanju dveh vprašanj: povezave skladnosti norm z oblikovanjem vrstniških skupin ter razlike med notranjim (osebnim) in zunanjim (vedenjskim) vidikom norm.V ta namen je potrebna analiza skladnosti znotraj skupin in med skupinami ocenjevalcev,ki ocenjujejo več objektov.Po teoretičnem pregledu obstoječih koeficientov rang-korelacije in konkordance je za analizo zbranih podatkov uporabljena ANACONDA (večskupinska analiza konkordance,analogna analizi variance,ki so jo v 1970-ih letih razvili Schucany in sodelavci).V okviru te metode je predlagana rešitev za doslej neobravnavani problem testiranja skupnega učinka posameznega dejavnika.Na koncu je prikazan postopek ocenjevanja intervala zaupanja za koeficient konkordance v primeru,da ničelna hipoteza ne drži.