Psychometric analysis of usability of the lower asymptote parameter in the graded response model for detecting socially desirable responding

Mutak, Augustin. (2018). Psychometric analysis of usability of the lower asymptote parameter in the graded response model for detecting socially desirable responding. Diploma Thesis. Filozofski fakultet u Zagrebu, Department of Psychology. [mentor Ljubotina, Damir].

Preview

PDF (Croatian) Download (569kB) | Preview

Abstract

Socially desirable responding (SDR) is one of the most common problems attributed to psychological tests measuring personality and attitudes. It is described as responding in a way that shows that a person has more favorable traits than she really has. The fact that item response theory has not been used as a tool for detecting SDR is surprising. Software that calculates a graded response model with the lower asymptote parameter (c) was developed in R under this thesis. A study aiming to discern whether the model with the c parameter fits the data better than the model without it was conducted. Also, diagnostic validity of the c parameter as a measure of SDR was assessed by comparing its value in groups that should theoretically differ in the extent of SDR and by correlating it with estimates of items' susceptibility to SDR. Analyses were performed on data downloaded from online repositories. The results show that models with the c parameters fit the data better than models without them and that groups' values of the c parameters differ in line with theoretical expectations. However, the parameters correlated poorly with estimates of susceptibility to SDR, a possible cause being estimators' low objectivity.

Actions (login required)

View Item