Likelihood ratio and score burden tests for detecting disease-associated rare variants

This paper presents two simple rare variant (RV) burden tests based on the likelihood ratio test (LRT) and score statistics. LRT is one of the commonly used tests in practical data analysis, and we show here that there is no reason to ignore it in testing RV associations. With the Bartlett correction, we have numerically shown that the LRT-based test can have a reliable distribution. Our simulation study indicates that if the non-null variants are as common as the null variants, then the LRT and score statistics have comparable performance to the C-alpha test, and if the former is rarer than the null variants, then they outperform the C-alpha test.