Department of Statistics, University Ca' Foscari, San Giobbe, Cannaregio 873, I-30121 Venice, Italy.
Biostatistics. 2010 Jan;11(1):127-38. doi: 10.1093/biostatistics/kxp042. Epub 2009 Nov 30.
Longitudinal data with binary and ordinal outcomes routinely appear in medical applications. Existing methods are typically designed to deal with short measurement series. In contrast, modern longitudinal data can result in large numbers of subject-specific serial observations. In this framework, we consider multivariate probit models with random effects to capture heterogeneity and autoregressive terms for describing the serial dependence. Since likelihood inference for the proposed class of models is computationally burdensome because of high-dimensional intractable integrals, a pseudolikelihood approach is followed. The methodology is motivated by the analysis of a large longitudinal study on the determinants of migraine severity.
在医学应用中，经常会出现带有二分类和有序分类结果的纵向数据。现有的方法通常是为处理短测量序列而设计的。相比之下，现代的纵向数据可能会产生大量特定于个体的连续观测值。在这个框架中，我们考虑了具有随机效应的多元概率比例模型，以捕捉异质性和自回归项来描述序列相关性。由于由于高维不可积积分，拟似然方法是一种计算上繁琐的方法，因此不适合对所提出的模型类进行似然推断。这种方法学的动机来自于对偏头痛严重程度决定因素的大型纵向研究的分析。
