This dissertation evaluates the effectiveness of various Differential Item Functioning (DIF) detection methods in multidimensional testing environments, particularly focusing on their application to polytomous data settings. The study aims to address the complexities that arise when traditional unidimensional approaches are used for inherently multidimensional constructs. Data for the study were gathered through both computer-simulated scenarios and real data collected via the shortened form of Emotional Skills and Competence Questionnaire (ESCQ-27). Four DIF detection methods were compared: Item Response Theory-Likelihood Ratio (IRT-LR), MIMIC with interaction, logistic regression with latent trait estimation, and logistic regression with observed total scores. The efficacy of these methods was assessed through their Type I error rates and statistical power (sensitivity). The findings revealed that logistic regression with latent trait estimation was most effective in identifying DIF in multidimensional polytomous data, efficiently detecting low DIF and demonstrating robustness against actual differences in group abilities and group sizes. Conversely, the MIMIC method with interaction tended to increase Type I error rates, indicating potential issues with false positives. Although simulated data supported the effectiveness of the logistic regression methods, the empirical results from the ESCQ-27 questionnaire emphasized the need for cautious interpretation and generalization due to the sample limitation in the empirical component of the study. The research highlights the necessity of using appropriate multidimensional DIF detection methods for complex constructs measured by psychometric instruments and recommends logistic regression with latent trait estimation for its superior performance in diverse testing scenarios. Future studies should focus on further developing and testing DIF detection methodologies to enhance their applicability and address existing limitations in different test situations and populations. This aim would improve the accuracy and reliability of psychometric evaluations across varied demographic groups.