0
$\begingroup$

I have two groups (healthy and disease, different sample sizes) of data, pooled from the individual subject's measurements of a certain region of the brain. The size of the brain region may be different from each individual and the sample sizes were different in these two groups. I have just carried out the ks.test using R on the data. They are not normally distributed.

As such, I ended up having 2500 samples for the healthy group and 1000 samples for the disease group.

I have plotted the histogram of these two groups (using normalized counts for y axis) and they appeared to be two different distributions, with overlapping. I would like to determine the cut-off value between two groups. What would be the correct way to go about it?

Improve this question
$\endgroup$
1
  • $\begingroup$ Follow the advice from @FrankHarrell in his answer. Much better than a cutoff is to describe and compare the full distributions for the two groups. $\endgroup$
    – EdM
    Jul 30, 2016 at 14:12

1 Answer 1

Reset to default
4
$\begingroup$

Forcing any continuous assessment to be classified using a cutoff/threshold is ultimately non-reproducible, arbitrary, and has very low precision. Plus if you use the data to select the cutoff you will need to bootstrap the whole process to get a reasonable measure of variability/stability.

Natura non facit saltus (Nature does not make jumps) -- Gottfried Wilhelm Leibniz

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.