Metrijske karakteristike instrumenata opisanih na malim uzorcima

In order to calculate metric characteristics of instruments applied to small samples, a canonical covariance analysis or quasicanonical analysis has been applied. Instrument reliability, described on a small sample of subjects, is defined as quasicanonical reliability in parallel or TAU equivalent forms model. Both an algorithm and a metric characteristics analysis program of a composite instrument have been made. This program calculates reliability in a quasicanonical model as a correlation between principal components of two forms. If an instrument is used to assess a small sample in two different temporal stages, defining the initial testing as the first, and the final testing as the second form, it is possible to assess stability of the measuring instruments. Besides quasicanonical reliability, this program also calculates the quasicanonical error as well as the validity and disriminativity coefficients for both forms. Standard reliability measure can be calculated as a correlation between Burt's components of the two forms. By appyling this model for analyzing metric characteristics of a questionnaire « Methods of a sighted guide », with the blind and the visually impaired, where the first form is consisted of independence characteristics, and the second form is made of motivation characteristics, satisfactory measuring features in a small sample of subjects have been obtained. Very high coefficients of quasicanonical reliability and congruency between validity coefficients imply the possibility of using these models in order to analyze metric characteristics of instruments when having small samples of subjects in research studies. In other words, the metric characteristics analysis can be efficiently used even in small samples of subjects if there is a possibility of explicit definition of two parallel forms.