Why False discovery Rate is same to raw p value in proc multtest?

HappySASUE — Fri, 02 Oct 2020 14:24:16 GMT

I use a simple example to illustrate my question. I want to check the fdr for a raw_palue, say 0.61, using the code below:

data a;

input Test$ Raw_P @@;
datalines;
test01 0.61
;

proc multtest inpvalues=a bon fdr ;
run;

p-ValuesTest Raw Bonferroni False DiscoveryRate1

0.6100

But if I use another dataset:

data a;
input Test$ Raw_P @@;
datalines;
test01 0.001 test02 0.001 test03 0.001
test04 0.02 test05 0.22 test06 0.59
test07 0.87 test08 0.61
;

proc multtest inpvalues=a bon fdr ;
run;

the output is

p-ValuesTest Raw Bonferroni False DiscoveryRate12345678

0.0010	0.0080	0.0027
0.0010	0.0080	0.0027
0.0010	0.0080	0.0027
0.0200	0.1600	0.0400
0.2200	1.0000	0.3520
0.5900	1.0000	0.6971
0.8700	1.0000	0.8700
0.6100	1.0000	0.6971

So the fdr for 0.61 become 0.6971?

Can anyone tell me why different dataset, with the same value, give me the different fdr?

Did I do something wrong?

Best,

Re: Why False discovery Rate is same to raw p value in proc multtest?

StatDave — Fri, 02 Oct 2020 14:58:09 GMT

See "p-value Adjustments" in the Details section of the MULTTEST documentation. The adjusted p-value for a particular test depends on the p-values of all of the other tests. If there is only one test, no adjustment can be done.

topic Why False discovery Rate is same to raw p value in proc multtest? in Statistical Procedures

Why False discovery Rate is same to raw p value in proc multtest?

Re: Why False discovery Rate is same to raw p value in proc multtest?