台大心理系

回首頁 演講訊息 108.12.11(三) 14:30 黃柏僩 副教授 〈Model Selection and Postselection Inference in Structural Equation Modeling〉
12/05/2019

108.12.11(三) 14:30 黃柏僩 副教授 〈Model Selection and Postselection Inference in Structural Equation Modeling〉

  • 演講時間: 108年12月11日(三) 14:30
  • 演講地點: N100
  • 講者: 黃柏僩 副教授(成功大學心理學系)
  • 演講主題: Model Selection and Postselection Inference in Structural Equation Modeling

Model selection is a popular strategy in structural equation modeling (SEM). In this talk, I will present a series of studies concerning the issue of model selection in SEM. First, the asymptotics of Akaike information criterion (AIC) and Bayesian information criterion (BIC) are discussed. The results show that both AIC and BIC asymptotically select a model with the smallest population minimum discrepancy function (MDF) value regardless of nested or non-nested selection, but only BIC can consistently choose the most parsimonious one under nested model selection. Second, a general framework for sparse estimation via penalized likelihood (PL) is presented. The framework integrates traditional specification search methods and modern regularization approaches in a unified manner. Under this framework, the sparsity pattern of a parameter matrix can be efficiently explored. The methodology can be easily implemented with R package lslx. Finally, the issue of postselection inference is discussed. In general, a model selection process destroys the asymptotic normality of estimators and makes the standard inference procedures invalid - which is probably a prime source of the reproducibility crisis in psychological science. Three valid postselection inference methods are accommodated for SEM, including data splitting, PoSI, and polyhedral methods. A simulation is conducted to compare the three valid methods with a naive procedure. The results show that the naive method often yields incorrect inference, and that the valid methods control the coverage rate in most cases with their own pros and cons.

回首頁 演講訊息 108.12.11(三) 14:30 黃柏僩 副教授 〈Model Selection and Postselection Inference in Structural Equation Modeling〉