The Mantel-Haenszel Statistic has been widely recommended for the statistical comparison of groups of observation of event times in which some observations may be censored. It is the test of most local power for detecting differences in failure rates, and has many applications in clinical trids and carcinogenesis experiments.
In order to show the result, In this paper, computation procedure is used the variance ratio (V/S) of Mantel-Haenszel statiseic (V) and indirect standardization statistic(S). Then survival distribution and censoring distribution determine the values of No/N. (N). presents the total number of death, N presents the total number at risk) According to this result, incresing survival distribution¢¥s V and censoring distribution V¢¥, No/M increases rapidely. (Survival distribution F(t)=e^vt, censoring distribution I(t)=e^-v¢¥t) But censoring distribution influenced No/N more sensitively. This means that censoring distribution influences more than survival distribution on the test statistic. We can apply this result to experimental design for toxicology study of clinical trials, and determine number of sample individuals at risk when we have prior information about v and v¢¥.
|