Article,
Performance of several variable-selection methods applied to real ecological data
Ecology Letters, 12 (10): 1061--1068 (October 2009)
DOI: 10.1111/j.1461-0248.2009.01361.x
Meta data
Tags
Users
Comments and Reviewsshow / hide
- @yourwelcome 8 years ago (last updated 8 years ago)This article has done a great disservice to the ecological modelling community by blurring the deep philosophical differences between the AIC and BIC. By its own simulations and practical analyses (low-dimensional, no interactions, sparsity), it constructs a situation in which one would not expect to find much difference between the AIC and BIC. Differences will only stronly manifest if one has an approximately infinite dimenionsal model, with the number of potential covariates or their order and number of interactions growing as sample-size grows.