help
help > RE: Multiple Comparisons Correction?
Jun 18, 2019 01:06 AM | Andrew Zalesky
RE: Multiple Comparisons Correction?
Hi Daniel,
If you are only running a few independent tests, I probably would not perform any correction and simply report the p-values for each test. Alternatively, if you are running say 10 or more tests, you could perform FDR.
With a -1/1 t-test, the p-value is one-sided. This means that you have an a priori hypothesis about the direction of the effect (i.e. positive correlation: +1; negative correlation: -1). With a F-test, the p-value is two-sided, meaning that you do not have any a priori hypothesis about the direction of the effect.
Andrew
Originally posted by Daniel Brennan:
If you are only running a few independent tests, I probably would not perform any correction and simply report the p-values for each test. Alternatively, if you are running say 10 or more tests, you could perform FDR.
With a -1/1 t-test, the p-value is one-sided. This means that you have an a priori hypothesis about the direction of the effect (i.e. positive correlation: +1; negative correlation: -1). With a F-test, the p-value is two-sided, meaning that you do not have any a priori hypothesis about the direction of the effect.
Andrew
Originally posted by Daniel Brennan:
Hello Andrew and others,
Do you suggest any multiple comparisons correction for p-values of individual NBS tests besides Bonferroni?
Also, what is the difference between testing for pos/neg correlation with 1/-1 t-test, and testing for any correlation with an F-test?
Thank you
Do you suggest any multiple comparisons correction for p-values of individual NBS tests besides Bonferroni?
Also, what is the difference between testing for pos/neg correlation with 1/-1 t-test, and testing for any correlation with an F-test?
Thank you
Threaded View
Title | Author | Date |
---|---|---|
Daniel Brennan | Jun 17, 2019 | |
Andrew Zalesky | Jun 18, 2019 | |