Advances and Applications in Statistics
Volume 39, Issue 1, Pages 1 - 23
(March 2014)
|
|
A TEST FOR MAIN EFFECTS WHEN OBSERVATIONS ARE RANDOMLY RIGHT CENSORED
A. Shanubhogue and D. P. Raykundaliya
|
Abstract: In this article, we generalize the Gore-test (Gore [8]), along the lines of Breslow [3] and Patel [16], for randomized block design when the observations are subject to arbitrary right censorship. The distributions of censoring variables are allowed to vary from block to block. The asymptotic distribution of the proposed statistic under the null hypotheses as well as Pitman type alternative is shown to be chi-square. The asymptotic power study is made when error distributions are logistic, Laplace, normal, Pareto of first kind, exponential, generalized exponential and Weibull. It is seen that the test performed well for most of the positive error distributions. |
Keywords and phrases: randomized block design, censoring, Pitman type alternative hypothesis, asymptotic chi-square, logistic, Laplace, normal, Pareto of first kind, exponential, generalized exponential, Weibull. |
|
Number of Downloads: 353 | Number of Views: 857 |
|