I have a distribution of shape sizes under two different (biological) conditions. From prior knowledge, I do expect there to be two populations. I fit each condition to a two-Gaussian mixture model. Here's the data:

Here's the data.

I want to know if the mean of the first gaussian (i.e. the left peak) is statistically different between the two conditions.

How could I perform a statistical test? I was thinking to do unpaired T-test, but use the mean and variance from the Gaussian fit, and estimate the N by counting all data points under each curve. Where the curves overlap, I would just divide the data according to the composition of the gaussian mixture model.

For example, if at area = 160, gaussian 1 is 60% and gaussian 2 is 40% of net mixture model (gaussian1+gaussian2), and I have 10 measurements, then I would have N=6 for gaussian 1 and N=4 for gaussian 2 at this area. This way I build up each gaussian with a finite number of observations. Then I would have all the parameters to calculate a unpaired t-test.

Has anyone come across a similar situation? How wrong is my proposed method? Please enlighten me!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.