The effect of response styles on data clustering: The case of PISA 2015

The attitude-achievement paradox is that the relationship between attitude and achievement in the country is positive, while it is negative among countries. Response style bias is considered one of the reasons for the attitude-achievement paradox and negatively affects the validity of statistical analysis results. In this regard, this study aims to ensure the content-based clustering of students from countries participating in PISA 2015 by correcting their response style bias and to explain the relationship between attitude and success within and between countries by determining the effect of response style bias on attitude-achievement paradox. The sample of the study, conducted in the correlational design, consists of 15-year-old students from 21 countries that participated in PISA 2015 survey and were selected among the countries with the attitude-achievement paradox. In this study, Enjoyment of Science, Achievement Motivation, and Science Self-Efficacy scales were used to determine the affective characteristics of students in science. In analyzing the data, the Correcting and Clustering Response Style Biased Data (CCRS) method was used to correct response style bias. Corrected scores, devoid of response style bias, are obtained using the CCRS method, and students are clustered based on these values. In addition, the K-means clustering method was also used over the uncorrected raw scores and compared with the CCRS method. To assess the significance of the difference between uncorrected and corrected science attitude scores, a related samples t-test analysis was performed for each country. The correlation values based on uncorrected and corrected scores were obtained in order to examine the effect of response styles on the relationship between attitude towards science and science achievement. According to the results obtained from the study, response style bias in all three scales in the study was corrected with the CCRS method, and content-based clusters were obtained. Moreover, response style-based clusters were obtained with the CCRS method. It has been observed that the response style-based and content-based clusters obtained by the CCRS method are different from each other. In the clustering obtained by the K-means method for all three scales, the students were not divided into clusters in terms of the measured trait, and similar results to the response style-based clustering were obtained. CCRS method was more successful than K-means clustering method for clustering individuals. For all three scales, it was concluded that the difference between the uncorrected and corrected attitude towards science score averages in many of the countries was significant. Intercountry attitude-achievement correlation values obtained from uncorrected and corrected scores were found to be negative for all three scales. The attitude-achievement correlation values between countries obtained through corrected scores are lower than the correlation values obtained from uncorrected scores. When all of the findings are evaluated together; it has been seen that one of the reasons for the attitude-achievement paradox is the response style bias, but the correction of the response style bias solely is not sufficient to eliminate this paradox; the response style bias can be corrected with the CCRS method, and clustering the students with the CCRS method in terms of the measured trait gives more valid results.