qiq15001
commented
5 years ago The results for mixture of normal distribution with/without using experimental data are different. I'm not pretty sure the result using experimental data is better. For simulated mixed gamma data, my function does not converge, the functions from existing package provide not very accurate results.
The normal assumption can be tested with the newly collected data points. With a sample size of 30, you would not have much power. You could test goodness of fit of skewed distributions such as gamma too. In the mixture modeling, you could model a mixture of gamma in addition to a mixture of normal.