chiragjp
commented
1 year ago On Dec 23, 2022, at 8:22 AM, Manh Thang @.***> wrote:
Dear Professor, I hope this one finds you well. So far, I am trying to understand how to calculate the false discovery rate by permutation. patel, 2019 # Unraveling exposure_Exposome wid.pdf https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_chiragjp_xwas-5Fwith-5Fnhanes-5Ftutorial_files_10296099_patel.2019.Unraveling.exposure-5FExposome.wid.pdf&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=Fkr4cKYb4VN6gRK-RJB393QKhtAj5_woms_Lxw8fSok&m=pSXi6Ptawv1gQA624opsXt9WH1uHRW7YkylWKAyZtzc5M1FTbt2tfIk3ZgQk4_5S&s=TTmMdTwsAH9kxenT6EQcAdjyE4kaCQ5R1-opoMOYUfI&e= There are some parts in your writing that I have not fully understood yet.
In Stage 2: Controlling for Multiple Hypotheses by Estimating the False Discovery Rate page 327, it is said that "The FDR is the ratio of the proportion of results that were called significant at a given level a in the null distribution and the proportion of results called significant from our real tests". As I understand, both the numerator and denominator are proportion. But line 11 in the pseudo-code said that "denominator <- sum(Pvalues < pvalue)" which is not a proportion. Which one should I follow?
That is correct; follow the former.
Also in page 327, it is said that "We use a significance level that corresponds to FDR of 5–10% to select associations". Could you please tell me what is the exact value of the significance level that corresponds to FDR of 5-10% and how to use that significant level? As I understand, for example, we chose the value of significance level as 0.0001. Then line 10 and line 11 in the pseudo-code could be numerator <- sum(nullPvalues < 0.0001)/numberPermutations denominator <- sum(Pvalues < 0.0001). Am I correct?
The value that corresponds to that FDR is the max pvalue that is lower than the FDR threshold.
What is the purpose of finding the minimum FDR? As I understand, we have P exposure variables. For each exposure variable, we will have one raw FDR. So we have P numbers of raw FDR. Then the minimum FDR we find the lowest FDR among the P number of raw FDR. Please correct me if I am wrong. Thank you for your time! — Reply to this email directly, view it on GitHub https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_chiragjp_xwas-5Fwith-5Fnhanes-5Ftutorial_issues_1&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=Fkr4cKYb4VN6gRK-RJB393QKhtAj5_woms_Lxw8fSok&m=pSXi6Ptawv1gQA624opsXt9WH1uHRW7YkylWKAyZtzc5M1FTbt2tfIk3ZgQk4_5S&s=MbExFsf58QtFuKFn-4dzLhd0PYOApjVzsGL3KIKEgKA&e=, or unsubscribe https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_ABFW2BRNCBZJ7JQ6TOIRS4TWOXGVXANCNFSM6AAAAAATH4SX7Y&d=DwMFaQ&c=WO-RGvefibhHBZq3fL85hQ&r=Fkr4cKYb4VN6gRK-RJB393QKhtAj5_woms_Lxw8fSok&m=pSXi6Ptawv1gQA624opsXt9WH1uHRW7YkylWKAyZtzc5M1FTbt2tfIk3ZgQk4_5S&s=cZj2L-ppeSCXj3Prpdy4DJN0T4HZ4ioB-ZVHA7cby3M&e=. You are receiving this because you are subscribed to this thread.
Dear Professor, I hope this one finds you well. So far, I am trying to understand how to calculate the false discovery rate by permutation. patel, 2019 # Unraveling exposure_Exposome wid.pdf There are some parts in your writing that I have not fully understood yet. 1) In Stage 2: Controlling for Multiple Hypotheses by Estimating the False Discovery Rate page 327, it is said that "The FDR is the ratio of the proportion of results that were called significant at a given level a in the null distribution and the proportion of results called significant from our real tests". As I understand, both the numerator and denominator are proportion. But line 11 in the pseudo-code said that "denominator <- sum(Pvalues < pvalue)" which is not a proportion. Which one should I follow? 2) Also in page 327, it is said that "We use a significance level that corresponds to FDR of 5–10% to select associations". Could you please tell me what is the exact value of the significance level that corresponds to FDR of 5-10% and how to use that significant level? As I understand, for example, we chose the value of significance level as 0.0001. Then line 10 and line 11 in the pseudo-code could be numerator <- sum(nullPvalues < 0.0001)/numberPermutations denominator <- sum(Pvalues < 0.0001). Am I correct?