r - Multiple testing methods -

r - Multiple testing methods -

i want simulate effect of different kinds of multiple testing correction such bonferroni, fisher's lsd, duncan, dunn-sidak newman-keuls, tukey, etc... on anova.

i guess should run regular anova. , take important p.values calculate using p.adjust. i'm not getting how p.adjust function works. give me insights p.adjust() ?

when running:

> p.adjust(c(0.05,0.05,0.1),"bonferroni") # [1] 0.15 0.15 0.30

could explain mean?

thank answer. kinda know bit of that. still don't understand output of p.adjust. i'd expect that...

p.adjust(0.08,'bonferroni',n=10)

... returns 0.008 , not 0.8. n=10 doesn't mean i'm doing 10 comparisons. , isn't 0.08 "original alpha" (i mean threshold i'd utilize reject null hypothesis if had 1 simple comparison)

you'll have read each multiple testing correction technique, whether false discovery rate (fdr) or family-wise error rate (fwer). (thanks @thelatemail pointing out expand abbreviations).

bonferroni correction controls fwer setting significance level alpha alpha/n n number of hypotheses tested in typical multiple comparing (here n=3).

let's testing @ 5% alpha. meaning if p-value < 0.05, reject null. n=3, then, bonferroni correction, split alpha 3 = 0.05/3 ~ 0.0167 , check if p-values < 0.0167.

equivalently (which straight evident), instead of checking pval < alpha/n, take n other side pval * n < alpha. alpha remains same value. so, p-values multiplied 3 , checked if < alpha = 0.05 example.

therefore, output obtain fwer controlled p-value , if < alpha (5% say), reject null, else you'd take null hypothesis.

for each tests, there different procedures command false-positives due multiple testing. wikipedia might start point larn other tests how right controlling false-positives.

however, output of p.adjust, gives in general multiple-testing corrected p-value. in case of bonferroni, fwer controlled p-value. in case of bh method, fdr corrected p-value (or otherwise called q-value).

hope helps bit.

r statistics