combine_results: Combine results of permutation test for keyness

In thmild/keyperm: Keyword Analysis Using Permutation Tests

Combine results of permutation test for keyness

Description

Combine results of two runs of keyperm() with output = "counts", possibly with different subsets of terms.

Usage

combine_results(results_1, results_2)

Arguments

results_1	Results from permutation test. Must be of class keyperm_results_counts (obtained by setting output = "counts" in keyperm())
results_2	Results from permutation test. Must be of class keyperm_results_counts and have the same scoretype as results_1.

Details

Results of two runs of keyperm() with output = "counts", i.e. objects of type keyperm_results_counts using can be combined using combine_results(). For this to make sense, scoretype needs to be the same in both results, but terms in both objects need not be the same.

There are at least two important uses of the function:

Parallelization: keyperm() is run several times with the same parameters on different cores, using parallel::mclapply() or a similar function.

Screening runs: keyperm() is first run using a small to medium number of permutations, but considering all terms. Terms with p-values clearly exceeding some reasonable significance threshold are then excluded, and keyperm() is run a second time with a (preferably) large number of permutations but using only the remaining terms. The results of both runs can then be combined into one object. The rationale behind this approach is that in many cases small p-values need to be determined with much greater accurary than larger ones far away from significance, especially if a correction for multiple testing is to be applied or the p-values are used for ranking (although they should not...).

Value

An object of class keyperm_results_counts

thmild/keyperm documentation built on Sept. 12, 2023, 12:25 a.m.