The investigation of measurement invariance PISA 2018 reading skills test according to countries and gender

The aim of the study is to examine the measurement invariance of the PISA 2018 reading literacy test across different countries and in terms of gender variable. Turkey, Singapore and the United States of America were selected to be used in the study among the countries participating in PISA 2018. The research was conducted on 24 different items belonging to 5 units in the reading literacy test. The units selected within the scope of the research were verified with confirmatory and exploratory factor analysis, and measurement invariance for countries and gender was examined for each unit. Measurement invariance between groups was analyzed by multi-group confirmatory factor analysis. As a result of analysis, structural invariance phase was ensured for five units selected, whereas metric invariance phase was not ensured for only one unit. For the unit not ensured in the metric invariance phase, it can be interpreted that item factor groups are equally distributed, but factor loadings, variances, covariances and error variances may differ. Two of the other four units ensured the scale invariance phase, while the other two units failed in the metric invariance phase. For units ensured the metric invariance phase, it can be interpreted that item and factor groups, as well as item factor loads, were similarly obtained according to the groups. In units with scale invariance; it can be reported that item and factor groups, item factor loadings and item constants were similar between groups. The fact that the scale invariance was ensured shows that the observed variable mean of the groups are comparable. According to the fit index obtained by gender with Turkey data; the stages of structural invariance, metric invariance and measurement invariance were ensured for all five units. It was observed that the three units ensured the stage of rigid invariance, two units failed at the stage of scale invariance. It can be interpreted that the observed averages of the three units between gender groups are comparable and that possible differences reveal gender differences. As the two units ensured the stage of scale invariance according to gender, it can be reported that there is no bias according to the items in these units and the differences obtained from the observed variables are due to the mean differences of the implicit variables.