On Estimating Good Reliability Coefficient when the Test is Split into Several Formats of Subtests and Standardizing the Raw Score, whose Distribution is Departed from Normality.

부문항이 분할된 고사에서 우량한 신뢰도 계수추경과 그 평가치 분포의 정규화

In this thesis. we estimated the good reliability coefficient ${\beta}$$\sub$k/ that is unbiased, consistent and more efficient than Cronbach's ${\alpha}$$\sub$k/ in splitting of a test into several formats of subtests and several properties of ${\beta}$$\sub$k/ are also represented. The tables of coefficients of skewness and kurtosis are represented to test the significance of departures from normality. We got the cumulative normal plots of z'from the distribution which is departed from normality using the Bock's approximation procedure and we finally enumerated the transformed standardized scores z'and a new raw score X' which enable us to proceed further evaluation procedures depending on our assessment policy.