Mixture modeling is useful when analyzing data where there is a heterogeneous population. In rare cases that mixture analysis yields results supporting that there is one latent class within the data. Nevertheless, this does not always mean that there is a genuine latent class in the sample analyzed. This paper has tested this idea with a series of simulations studies and demonstrated in various conditions that mixture Rasch model in general gives support to solutions of multiple-class solution, though the simulation data is indeed generated under a uni-dimensional, one-class specification only.
Comment:
1. This article brings alert to researchers applying mixture modeling techniques that they should be aware of the impact of “spurious” latent classes when interpreting a result of multiple-class solution.