I am also having the same problem as others have had when the p-values are extremely low. For example, I'm trying to compare my sQTL finding with those from GTEx and wanted to calculate the pi1 statistic to see the proportion of true positive hits in my data. The problem seems to be that the p-values are extremely low; for example: the lowest p-value in my dataset is 3.07e-87 and the highest p-value is 4.06e-4. Therefore, none of my p-values are higher than the upper limit of the lambda vector. I was wondering if modifying the default parameters (lambda = seq(0.05, 0.95, 0.05)) for lambda is an appropriate course of action to solve the problem of handling very low p-values. If this is the case, how would I go about modifying lambda in terms of new parameters?
Hello,
I am also having the same problem as others have had when the p-values are extremely low. For example, I'm trying to compare my sQTL finding with those from GTEx and wanted to calculate the pi1 statistic to see the proportion of true positive hits in my data. The problem seems to be that the p-values are extremely low; for example: the lowest p-value in my dataset is 3.07e-87 and the highest p-value is 4.06e-4. Therefore, none of my p-values are higher than the upper limit of the lambda vector. I was wondering if modifying the default parameters (lambda = seq(0.05, 0.95, 0.05)) for lambda is an appropriate course of action to solve the problem of handling very low p-values. If this is the case, how would I go about modifying lambda in terms of new parameters?
Thank you, Nelson