Department of Biostatistics and Computational Biology, University of Rochester, Rochester, NY, United States of America.
Department of Biomedical Engineering, University of Rochester, Rochester, NY, United States of America.
PLoS One. 2022 Apr 5;17(4):e0265471. doi: 10.1371/journal.pone.0265471. eCollection 2022.
When dealing with longitudinal data, linear mixed-effects models (LMMs) are often used by researchers. However, LMMs are not always the most adequate models, especially if we expect a nonlinear relationship between the outcome and a continuous covariate. To allow for more flexibility, we propose the use of a semiparametric mixed-effects model to evaluate the overall treatment effect on the hemodynamic responses during bone graft healing and build a prediction model for the healing process. The model relies on a closed-form expectation-maximization algorithm, where the unknown nonlinear function is estimated using a Lasso-type procedure. Using this model, we were able to estimate the effect of time for individual mice in each group in a nonparametric fashion and the effect of the treatment while accounting for correlation between observations due to the repeated measurements. The treatment effect was found to be statistically significant, with the autograft group having higher total hemoglobin concentration than the allograft group.
当处理纵向数据时，研究人员通常会使用线性混合效应模型（LMMs）。然而，LMM 并不总是最合适的模型，特别是如果我们预期结果与连续协变量之间存在非线性关系。为了提供更大的灵活性，我们建议使用半参数混合效应模型来评估在骨移植物愈合过程中的血流动力学反应的整体治疗效果，并建立愈合过程的预测模型。该模型依赖于封闭形式的期望最大化算法，其中使用套索型程序来估计未知的非线性函数。使用这种模型，我们能够以非参数方式估计每个组中个体小鼠的时间效应，以及在考虑到由于重复测量引起的观测相关性的情况下的治疗效果。治疗效果具有统计学意义，自体移植物组的总血红蛋白浓度高于同种异体移植物组。
