Item Response Theory Model for Understanding Item Non- Response in Ghanaian Surveys

Abstract

This paper explores four Item Response Theory (IRT) models to determine the most appropriate forunderstanding item non-response. The selected IRT model was used to identify among five categories of surveyquestions, the most difficult to answer by respondents and determine the underlying mechanism behind missingdata which is defined to include ‘don’t know’ answers. A questionnaire data on Ghana collected in the fifth waveof the World Values Survey was implored. All items were dichotomously scored. Missing or ‘don’t know’responses were assigned a 0 score whiles answered items were assigned a 1 score. The four IRT models thatwere explored included both the constrained and unconstrained versions of the Rasch model, the two parameterlogistic model (2-PLM), and the three parameter logistic model (3-PLM). The unconstrained Rasch modelemerged as the most appropriate model for understanding item non-response. It was observed that, incomerelated questions had the highest difficulty parameter, hence the most difficult category of survey questions toanswer. It was also found that, if an individual does not answer a survey question or give a ‘don’t know’ answer,it is not only due to the question’s difficulty but also because the respondent doesn’t want to answer.Keywords: Item Non-response, Item Response Theory (IRT), World Values Survey (WVS).