The Study of Variance Components in an MC Achievement Test with Parallel Items Using G-Theory

The variability in a test can be estimated by variance component models for each level of the hierarchy. In an MC achievement test of structure 30 persons (p) each took a different pair of items. Its design had items nested within persons, as in i:p, that is, one-facet , nested design. The researcher used ANOVA to determine the magnitudes of the diverse variance components. Using G- and D- studies, most of the variance in achievement scores was attributable to the residual variance component . In this residual term the item effect, the person-by-item interaction, and remaining sources of systematic and unsystematic variation not estimated in the research are all confounded. Since each person received a different item, it was not possible to whether items diff ered in difficulty level (item effect), or whether some items were particularly difficult for some participants (person-by-item interaction effect). The results showed that the malfunctioning on an item for each person could have occurred since the item wa sparticularly difficult or since it was particularly difficult for remaining not calculated sources of variation, such as the person accidentally marking the wrong answer on theanswer sheet

Keywords:

Variance Component , MC , Achievement Test , Parallel Items , G-Theory