Research Seminar
Identification Issues in PsychometricsErnesto San MartinDepartment of Statistics & Measurement Center MIDE, UC Santiago de Chile | |
| Abstract: | Parameter identification is a crucial issue when we are evaluating if a model has an empirical meaning. This problem becomes more relevant when a statistical model is obtained through a hierarchical construction, as typically done in psychometrics. The presentation is divided into three parts: 1) Can the identification of a statistical model after integrating out the latent variables be obtained from the identification of the model given the underlying hierarchy? In the IRT- and SEM-literature, it is typically assumed that this rule works. We aim at understanding the structure of this implication and will give an illustration for a fullu Bayesian discrete model. 2) A general result leading to identification of the Latent Class Models (LC) will be discussed. The result is based on a procedure introduced by Lazarsfeld (1950) and developed by Anderson (1954) and Madansky (1960). We "clean" the initial procedure and we prove that it is useful to obtain the identification of the parameters of interest in a LCM. An illustration will be given. 3) It is shown how the identification strategy used in the LCM case can be applied for studying the identification of the Multiple Classification Latent Class Models (MCLCLM). An alternative proof is discussed for the case of one latent class and at least three items. As a consequence of the identification analysis, it will be shown that the MCLCM is a particular case of the LCM, and not the converse as widely stated in the literature. |
| Date: | Tue Oct 28, 12:15 pm - 1:15 pm |
| Place: | room 00.60 (Department of Psychology, Tiensestraat 102, 3000 Leuven) |