Problem Statement: Student achievement is considered an indicator of the quality of education, and achievement tests are applied to assess student achievement. International tests are adapted into different languages and cultures with the aim of assessing student achievement on an international level and comparing the achievements of different countries. In our country, a number of tests at the national and international levels are conducted to assess student achievement. One of the tests conducted in our country is called Trends in International Mathematics and Science Study (TIMSS). Countries structure their curricula and education policies based on the results of these studies. However, in order for these comparisons to be meaningful, the constructs measured by the tests should be equivalent. When the relevant literature was examined, it was observed that the number of studies on cross-cultural invariance in Turkey was low and that these studies did not involve TIMSS 2011.
Purpose of the Study: The purpose of this study was to examine the measurement invariance of TIMSS 2011 mathematics test in terms of different cultures.
Method: Aiming at examining the intercultural measurement invariance of the TIMSS 2011 mathematics test, this is a survey model that tries to describe an existing situation as it is. The study sample was composed of 1,987 fourth graders from Turkey, England, Japan and the USA. This study was conducted on the data obtained from the TIMSS 2011 mathematics test. Model invariance was examined through multi-group confirmatory factor analysis. LISREL 8.80 for Windows software was used for performance of data analysis.
Findings and Results: The study of measurement invariance was conducted in four steps. It was found that the proposed model was confirmed for all countries, and configural invariance was ensured in the first step, while metric invariance was not ensured in the second step. Therefore, we did not start the scalar invariance or strict invariance analyses. After this step, metric invariance was tested through binary and trilateral combinations in order to determine in which country the invariance was collapsed. It was found that the reason why the metric invariance wasn’t ensured was that it was not sourced from only one country.
Conclusions and Recommendations: According to the findings, the invariance across four countries was ensured only in the configural invariance step. Therefore, the items causing the model not to have measurement invariance can be determined, as well as whether the items demonstrated DIF across groups. The items determined to demonstrate DIF can be examined in terms of bias of sources, depending on the expert opinions.
Keywords: Measurement invariance, Multiple-group confirmatory factor analysis, Structural equation modeling