Significance of PLINK epistasis p-values
I performed an epistasis analysis on PLINK using the following options: --epistasis --epi1 1
The output from PLINK reports tests significant at a p-value (asymptotic p-value) <1e-4 by default. Since I am not performing a GWAS and I only have 400 SNPs, I changed the output to include all tests.
Do I therefore consider all interaction tests smaller than p<1e-4 as significant or stick to the 'usual' p<0.05?
I performed a Bonferroni correction and nothing was significant. Therefore I am confused on how to interpret the PLINK output p-values.
• 4,922 views
•
link
updated
by
Ram
•
written
by
viavialin •
•
written
by
viavialin •
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
Weird results in PANTHER overrepresentation test - Bonferroni count:1 is this possible?
written by annamariabugaj •
-
Setting a threshold for FDR correction
written by kelly.wang135
-
t-test output from TIGR MeV
written by kashyap_path10 •
-
PANTHER gene ontology - is the bonferroni correction important?
written by sbrown669 •
-
Cuffdiff not performing tss, cds, splicing, promoter or CDS tests
written by BioinfGuru
-
HOMER tool's HiC interactions
written by hiren.knowledge •
-
Calculation A Corrected P-Value After Testing Or Setting A Uncorrected P-Value Cut Off
written by Floris Brenk
-
Fdr Q Values Calculations Using Q-Value Package
written by viavialin •
-
Using --Logistic --Interaction --Condition To Search For Epistasis In Plink, And Vif Issues
written by James Bullock •
-
Anova + Tukey: Multiple Testing Correction
written by Rossella •
You should read up on http://en.wikipedia.org/wiki/Multiple_comparisons_problem I would NOT go with the p<0.05 cutoff. Bonferroni correction might be a bit too conservative. Anything significant if you check FDR? If not, you might be out of luck.
try LAMP algorithm or lamplink tool directly;