|
Abstract:
|
In many situations it may be desirable to combine the results of several independent tests for a hypothesis of interest. The reason for combining tests may be that some data are lost or not conveniently available, or that the individual experiments are too dissimilar to combine. The combined test may be a function of the attained significance levels, say P1, P2,., Pk, of k independent tests or a function of their corresponding test statistics T1, T2, ., Tk. Combining the attained significance levels leads to a nonparametric test based on the Chi Square distribution. However, a test based on a function of the test statistics requires knowledge of the distribution of each Ti, as well as the distribution of the function of the statistics. In the present paper the authors discuss the above mentioned techniques of combining tests when testing the equality of means.
|
