Barr S C, O'Neill T J
Department of Statistics and Econometrics, Australian National University, Canberra ACT.
Biometrics. 2000 Jun;56(2):443-50. doi: 10.1111/j.0006-341x.2000.00443.x.
The analysis of group truncated binary data has been previously considered by O'Neill and Barry (1995b, Biometrics 51, 533-541), where the analysis assumed that responses within each group were independent. In this paper, we consider the analysis of such data when there is group-level heterogeneity. A generalized linear mixed model is hypothesized to model the response and maximum likelihood estimates are derived for the truncated case. A score test is derived to test for heterogeneity. Finally, the method is applied to a set of traffic accident data.
奥尼尔和巴里（1995b，《生物统计学》51卷，533 - 541页）之前曾考虑过对分组截断二元数据进行分析，其中分析假设每个组内的响应是独立的。在本文中，我们考虑当存在组水平异质性时此类数据的分析。我们假设一个广义线性混合模型来对响应进行建模，并针对截断情况推导最大似然估计。推导了一个得分检验来检验异质性。最后，将该方法应用于一组交通事故数据。
