Permutation tests are an interesting and conceptually simple alternative to traditional tests when the required distributional assumptions (typically, Gaussian assumptions) are likely to be violated. We start with three motivational examples explaining how to construct permutation and sampled permutation tests, and provide empirical evidence of the way they outperform their traditional competitors onseveral counts: validity, unbiasedness, and power. We then show how permutation tests follow from more traditional Neyman alpha-structure arguments, and discuss some asymptotic results establishing that extended validity, in this context, does not imply any loss of asymptotic efficiency.
|Title of host publication||Encyclopedia of Environmetrics, 2nd Edition|
|Editors||W. Piegorsch, A. El Shaarawi|
|Number of pages||3510|
|Publication status||Published - 2012|