Abstract:
Using the conditional likelihood, a model-specific score test can be derived under a given genetic model to test genetic association for case-parents triad family data. When the underlying genetic model is correctly specified, the score test is most powerful. However, it can lose substantial power when the model is misspecified. Several robust tests have been proposed to deal with the problem, such as the maximum test statistic, the maximin efficiency robust test, and the constrained likelihood ratio test. These tests have been shown to be robust against model misspecification compared with those model-specific score tests, but they are either time-consuming in computation or not sufficiently high in power robustness under some situations. In this study, a data-driven procedure is proposed to construct two adaptive robust genetic association tests WMERT and WMAX . The WMERT is simple in calculation and has fairly high power robustness. The empirical power of WMAX is quite stable and close to those of the model-specific score tests. The two proposed tests should be beneficial to practical genetic association studies. A real dataset consisting of neural tube defect triad families is used for illustration of the methods. R-scripts are also provided for numerical calculation of the proposed methods in practical studies.
摘要:
使用条件似然,可以在给定的遗传模型下得出特定于模型的分数检验,以检验病例-父母三合会家庭数据的遗传关联。当正确指定了基础遗传模型时,分数测试是最强大的。然而,当模型被错误指定时,它可能会失去大量的力量。已经提出了几个稳健的测试来处理这个问题,例如最大检验统计量,最大效率稳健测试,和约束似然比检验。与那些特定于模型的分数测试相比,这些测试已被证明对模型错误规范具有鲁棒性,但是在某些情况下,它们要么在计算中耗时,要么在功率鲁棒性方面不够高。在这项研究中,提出了一个数据驱动的程序来构建两个自适应稳健遗传关联测试WMERT和WMAX。WMERT计算简单,具有相当高的功率鲁棒性。WMAX的经验能力非常稳定,与特定于模型的分数测试接近。提出的两个测试应该有利于实际的遗传关联研究。由神经管缺陷三元组族组成的真实数据集用于说明该方法。还提供了R脚本,用于实际研究中提出的方法的数值计算。