Assessing Dimensionality of Latent Structure Underlying Dichotomously Scored Item Response Data

Abstract: 

Assessing Dimensionality of Latent Structure Underlying Dichotomously Scored Item Response Data

The purpose of this study is to investigate the statistical behavior of different analytical procedures recommended in the literature to assess the dimensionality of latent structure underlying dichotomous data. The project has three subsequent studies. First, the sampling behavior of decisions made by different analytical procedures regarding the dimensionality of dichotomous data will be investigated using empirical datasets. Second, model parameters will be estimated from empirical data for one-, two-, three-, and four-dimensional common factor models, and these parameters will be used to generate multidimensional dichotomous response data. The study will compare the bias and accuracy of decisions regarding the dimensionality suggested by different procedures under simulated data. Third, various variables such as test length, sample size, and latent structure will be manipulated, and more detailed analysis will be conducted under different simulation conditions for bias and accuracy of these procedures in assessing dimensionality of dichotomous data.

Group name: 
davenpor