Avtor/Urednik		Lapajne, Zdenko
Naslov		Delitev testa na postavke in ocena zanesljivosti
Tip		članek
Vir		Psihol obz Ljubl
Vol. in št.		Letnik 6, št. 3
Leto izdaje		1997
Obseg		str. 5-14
Jezik		slo
Abstrakt		The item is a fundamental concept in either classical or modern test theory but it is not defined inside the theory. An example is given of an "item", used for final examination in secondary education, that could consist of 4. 14 or even 32 items if the linguistic problem is considered to be at the level of word, syllable of phoneme. Since the number of items k appears in all formulae for reliability estimates, it could be expected, that higher reliability estimates will follow when the same test is divided into more items. IEA Reading Literacy Test data for 9-year Slovenian students in 1991 (N = 3.297) were studied to test this hypothesis. Individual 66 items were summarized into equal "compound items" of sums, consisting of 2, 3, 6, 11, 22, or 33 original items, chosen at random. The test was divided into 15 subtests also. according to 15 texts. The results show that compound items have higher item-total correlations, but the resulting reliability estimates are quite the same with alpha around .93. Reliability, estimated by correlating summation scores for 15 text parcels, was slightly lower: .88. Reasons for obtained resultst are discussed in some detail.
Deskriptorji		PSYCHOMETRICS REPRODUCIBILITY OF RESULTS