Abstract: | Argues that although E. F. Alf and N. M. Abraham's (see record 1972-22071-001) conclusion that for small samples, randomization tests may be considerably less powerful than the corresponding parametric tests is valid, it is applicable only to those experimental situations where a population has been randomly sampled. Where there is random assignment to treatments but not random sampling, as is frequently the case in psychological experimentation, power comparisons of parametric and randomization tests are inappropriate because parametric tests are then valid only to the extent that they approximate randomization tests. It is concluded that in experiments employing random assignment but not random sampling, one should use randomization tests, if possible, whether the sample is large or small. (PsycINFO Database Record (c) 2010 APA, all rights reserved) |