So I have a lot of known disease and normal data values from a test I am devising. I have run ROC on the unaltered data and gotten values etc but the data is heavily skewed. Normals sit close to 0 and disease have a mix of close to 0 off into the 50's.

Log^10 normalises the data for stats.

Can the Log data be used in ROC analysis and is it appropriate?

Thanks :)

More Stephen Reynolds's questions See All
Similar questions and discussions