Question: Not getting Adjusted p-values from multtest package
0
3.4 years ago by
elhananby10
elhananby10 wrote:

Hello everybody,

I'm trying to run an analysis on methylation percentage using the multtest package, but I'm running into a problem.

I have a matrix where each columns is a subject, and each row is a number between 0-100, corresponding to percentage of methylation at that specific loci on the genome. Further, i have a vector containing the groups coding, so '0' represent a control and '1' a case.

I'm trying to run a t-test between the two groups (cases and controls) using the mt.maxT function, and while I'm getting the t-values and raw-p-values, all of the values in the adjusted p-value (adjp) are 1.

I'm trying to run the function as so:

mt.maxT(methyl.matrix, coding_groups, test = "t", side ="abs", fixed.seed.sampling = "y", B = 1000)

I've tried to run the raw p-values through the mt.rawp2adjp function, but then I'm getting an adjusted p list of 1s and 0s.

Furthermore, when trying to run the analysis using the MTP function, i'm getting this error:

Only one unique value in bootstrap sample for first group. Cannot calculate variance. This problem may be resolved if you try again with a different seed.

This error remains even after removing all missing (NA) values list-wise.

Any help will be appreciated,

Thanks.

modified 3.4 years ago by kpollard110 • written 3.4 years ago by elhananby10
1
3.4 years ago by
kpollard110
United States
kpollard110 wrote:
It sounds like you have a lot of hypotheses and that after multiple testing adjustment all non-zero unadjusted p-values become 1s, which is probably the correct output of the maxT procedure. It is designed to control family-wise error, so will increase p-values a great deal when you have many tests. The function mt.rawp2adjp with proc=�BH� (or �BY�, �ABH�, or �TSBH�) will allow you to control FDR, which might be more satisfying in your scenario. This still may produce many 1s if you have a large number of tests and there are no significant differences. Best, Katie On Nov 26, 2015, at 4:59 AM, elhananby [bioc] <noreply@bioconductor.org> wrote: > Activity on a post you are following on support.bioconductor.org > User elhananby wrote Question: Not getting Adjusted p-values from multtest package: > > > Hello everybody, > > I'm trying to run an analysis on methylation percentage using the multtest package, but I'm running into a problem. > > I have a matrix where each columns is a subject, and each row is a number between 0-100, corresponding to percentage of methylation at that specific loci on the genome. Further, i have a vector containing the groups coding, so '0' represent a control and '1' a case. > > I'm trying to run a t-test between the two groups (cases and controls) using the mt.maxT function, and while I'm getting the t-values and raw-p-values, all of the values in the adjusted p-value (adjp) are 1. > > I'm trying to run the function as so: > > mt.maxT(methyl.matrix, coding_groups, test = "t", side ="abs", fixed.seed.sampling = "y", B = 1000) > I've tried to run the raw p-values through the mt.rawp2adjp function, but then I'm getting an adjusted p list of 1s and 0s. > > Furthermore, when trying to run the analysis using the MTP function, i'm getting this error: > > Only one unique value in bootstrap sample for first group. Cannot calculate variance. This problem may be resolved if you try again with a different seed. > This error remains even after removing all missing (NA) values list-wise. > > > Any help will be appreciated, > > Thanks. > > > Post tags: multtest, methylation, adjusted pvalue > > You may reply via email or visit Not getting Adjusted p-values from multtest package >