FDR column, in T-test output

Help
2011-11-02
2013-05-20
  • Haritz Irizar
    Haritz Irizar
    2011-11-02

    Hi.

    In a permutation-based t-test analysis output, to what parameter the False Discovery Rate column corresponds to?

    It`s obvius it doesn't correspond to the FDR value itself, as the FDR is an only value for all the analysis (the estimated proportion of false positives) and not a parameter calculated for each gene. So, what's that exactly? A corrected p-value for the gene?

    Thanks for your attention.

     
  • Hello,

    The false discovery rate column corresponds to the false discovery rate given a cutoff corresponding to that gene's p-value.
    In other words, if you had run the analysis with a cutoff equal to that gene's p-value, the estimated proportion of false positives would be the value seen in the False Discovery Rate column.

    Hope that helps,
    Dan

     
  • Haritz Irizar
    Haritz Irizar
    2011-11-03

    Hi Dan,

    Thanks for your answer.

    The problem is that after setting the threshold in p<0.01; and FDR<0.15 with 15000 permutations, in the output, the highest values of p-values and FDR of the significant genes don't correspond to thresholds I established (they vary between 0.00004<p<0.003 and 0.022<FDR<0.0026) and this has happenned to me several times. I'd like to understand that.

    Thanks again,

    Haritz