Fixed-effect variable selection in linear mixed models using R2 statistics
Fixed-effect variable selection in linear mixed models using R2 statistics
复制标题
DOI：
10.1016/j.csda.2007.06.006
复制
发表时间：
2008-01-10
影响因子：
1.8
通讯作者：
Edwards, Lloyd J.
中科院分区：
文献类型：
--
作者：
Orelien, Jean G.;Edwards, Lloyd J.
在线性混合模型（LMM）中，已经提出了几种R-2统计量来评估固定效应的拟合优度。然而，这些统计数据的性能还没有得到充分证明，无论是分析或通过模拟。我们报告的模拟结果，以评估这些统计选择最简约的模型的能力。将来自完整模型的R-2统计量与去除固定效应协变量的其他模型进行比较。还将完整模型与过度拟合模型进行了比较，过度拟合模型包括与结局无关的其他协变量。所有模型比较涉及相同的随机效应。在本文中，我们表明，涉及残差的R-2统计量是无法充分区分正确的模型和一个重要的固定效应协变量被省略，如果残差的预测值的计算包括随机效应（称为条件R-2统计量）。然而，如果从导致残差的预测值的计算中排除随机效应，则这些R-2统计量（称为边际R-2统计量）能够选择最简约的模型。Xu [2003]提出的其他R-2统计量。测量线性混合效应模型中的解释变异。中央集权主义者Med.22（22），3527-3541]表现不佳，因为从完整模型到简化模型，这些统计值几乎没有变化。(C)2007 Elsevier B. V.保留所有权利。
In the linear mixed model (LMM), several R-2 statistics have been proposed for assessing the goodness-of-fit of fixed effects. However, the performance of these statistics has not been fully demonstrated either analytically or through simulations. We report results of simulations to asses the ability of these statistics to select the most parsimonious model. R-2 statistics from a full model were compared to other models in which fixed-effect covariates were removed. The full model was also compared to an overfitted model that included additional covariates not linked to the outcome. All models compared involved the same random effects. In this paper, we show that R-2 statistics that involve the residuals are unable to adequately discriminate between the correct model and one from which important fixed-effect covariates are omitted if the computation of the predicted values for the residuals included the random effects (referred to as conditional R-2 statistics). However, if the random effects are excluded from the computation of the predicted values that lead to the residuals, these R-2 statistics (referred to as marginal R-2 statistics) are able to select the most parsimonious model. Other R-2 statistics that have been proposed by Xu [2003. Measuring explained variation in linear mixed effects models. Statist. Med. 22(22), 3527-3541] performed poorly in that there was little variation in the value of these statistics from a full model to a reduced model. (C) 2007 Elsevier B.V. All rights reserved.