Testing for Serial Correlation in Regression with Missing Observations
针对静态时间序列回归中因缺失数据导致的残差序列相关性检验问题，提出了基于条件似然和无条件似然的得分检验方法，并推导了检验统计量的渐近分布，可用于比较不同检验的效率。
SUMMARY In order to test for serial correlation in residuals for static time series regression in the presence of missing data, the score principle is applied both to the likelihood conditional on the observation times, and to an unconditional form of likelihood. Asymptotic distributions of the test statistics are established, under both the null hypothesis of no serial correlation, and sequences of local, correlated, alternatives, enabling analytic comparison of efficiency.