Model population analysis in model evaluation

• Published in: Chemometrics and intelligent laboratory systems Vol. 172; pp. 223 - 228
• Main Authors: Deng, Baichuan, Lu, Hongmei, Tan, Chengquan, Deng, Jinping, Yin, Yulong
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.01.2018
• Subjects: Model comparison; Model evaluation; Model population analysis; Model selection; Model evaluation; Model comparison; Model population analysis; Model selection
• ISSN: 0169-7439; 1873-3239
• DOI: 10.1016/j.chemolab.2017.11.016

Model evaluation plays a central role in chemical modeling. Model population analysis (MPA), a general framework for designing new types of chemometrics algorithms, has shown its advantage in the field of model evaluation. The core idea of MPA is to statistically analyze the outputs of randomly generated sub-models to extract interesting information from the data. One of the most obvious characteristics of MPA-based methods is that they use multiple models instead of a single model for model evaluation. In this review, we described the concept of MPA, and then discussed the application of MPA in model evaluation, including the relationship between MPA and cross-validation, model comparison, randomization tests, model stability, variable importance and sum of rank differences. Finally, we prospected the potential application of MPA in model evaluation. •Model population analysis (MPA) is a general framework for chemometrics algorithms.•We reviewed the development of MPA in model evaluation.•We prospect the potential application of MPA in model evaluation.