compensatePValue.Rd**OBSOLETED**
compensatePValue compensates a p-value for the number of tests
performed.
compensatePValue(p, n = length(p), method = c("bonferroni", "sidak", "holm-bonferroni"), r = 0)
| p | p-value(s) to be compensated |
|---|---|
| n | number of tests made |
| method | compensation method, 'bonferroni' is conservative, does not assume independence of the tests, and controls for the family-wise error (FWE), 'sidak' is more liberal, and controls for FWE only under the assumption of independence of the tests performed, 'holm-bonferroni' is an iterative procedure controlling for FWE |
| r | optional mean "correlation between the tests made", the extreme value of 0 leads to full compensation, the extreme value of 1 leads to no compensation (as all the tests are considered equal). See the SISA help for explanation. Note that the FWE is controlled in the strict sense only for 'r=0'. |
Compensated p-value. It equals min(1,p*n) for the 'bonferroni'
method, 1-(1-p)^n for the 'sidak' method, and the result of the
iterative procedure for the method of 'holm-bonferroni'.
SISA, http://www.quantitativeskills.com/sisa/calculations/bonhlp.htm Sture Holm, A Simple Sequentially Rejective Multiple Test Procedure, Scand J Statist 6: 65-70, 1979
deprecated, not working properly, use p.adjust instead
# demonstrate the difference between Bonferroni and Sidak: compensatePValue(.025,2,method='bonferroni')#> [1] 0.05compensatePValue(.025,2,method='sidak')#> [1] 0.049375# demonstrate the iterative Holm-Bonferroni method: #compensatePValue(c(.01,.02,.04,.05),method='holm-bonferroni') #compensatePValue(c(.05,.04,.01,.02),method='holm-bonferroni') # use 'stats::p.adjust' instead