High specificity in a ROC model with the lroc command

[Maria Santos]

I am new to the ROC analysis, but I need to learn for my biomakers analysis. I am trying to see if some biomarkers are capable of discriminating my ppopulation between two groups. For this, I am using the following commands:

1º logistic depvar predvar
2º lroc
3º lstat

I guess my data are too unbalanced cause I'm getting very high specificity for my model (some are 100%), but very low sensibility (even 0%). I saw that I need to do resampling or bootstraping, but I have no idea how to do it. Could somebody help me somehow?

[Rich Goldstein]

it is not clear how resampling of any type (including bootstrap) would help here (where did you find that suggestion?) and I, personally, doubt that it will help ; the usual situation where data are highly imbalanced is to change the cutoff from its default of 0.5 - however, what to change it too depends heavily on substantive knowledge and I don't have that here - as a default, you could either use the actual proportions in the data, or in the population, or some function of the predicted probabilities - but what you choose depends heavily on knowledge we don't have an you don't provide; further, any "simple" suggestion such as I gave examples of is unlikely to persuade others without some form of substantive argument and that, again, requires knowledge not provided in the question