摘要:Reliability of several Slovenian Matura examinations in years 1996-1999 is reviewed,mostly from the viewpoint of the classical test theory.Results of confirmatory factor analyses had shown that coefficient α is not the optimal reliability measure for such tests.Therefore,reliabilities were computed by means of minimum rank factor analysis.Reliabilities of some exams were not satisfactory,especially in cases where the proportion of objectively scored items in the exam was low.Further,changes of reliability through time and reliability of the complete examination score is discussed.Additionally,reliability estimation in the frame of item response theory is illustrated.Finally,some suggestions are made about possible improvements of reliability of exams and its assessment.
其他摘要:V prispevku obravnavamo zanesljivost nekaterih maturitetnih izpitov v obdobju 1996-1999,večinoma z vidika klasične testne teorije.Rezultati konfirmatornih faktorskih analiz so pokazali,da priljubljeni koeficient α zaradi neveljavnosti predpostavk ni najboljša mera zanesljivosti za tovrstne preizkuse.Zato smo zanesljivost računali z metodo faktorske analize najmanjšega ranga.Izkazalo se je,da zanesljivost nekaterih preizkusov ni zadovoljiva,zlasti v primerih,kjer je majhen delež objektivno ocenljivih nalog.V prispevku obravnavamo še spreminjanje zanesljivosti preizkusov skozi čas ter ocenjevanje zanesljivosti celotnega maturitetnega izpita,ob konca pa ilustriramo še ocenjevanje zanesljivosti po teoriji odgovora na postavko.Prispevek končujemo z nekaj praktičnimi nasveti,kako bi lahko izboljšali ocenjevanje zanesljivosti maturitetnih izpitov in njihovo zanesljivost samo.
关键词:reliability;Matura;Cronbach's alpha;factor analysis;item response theory
其他关键词:zanesljivost;matura;Cronbachova alfa;faktorska analiza;teorija odgovora na postavko