Hi Jesus,
I think I have done something similar to what you are describing in Relion.
I did a classification for 75 iterations into 3 classes and then looked at the class "commitment" in last 16 iterations for each particle. I defined "commitment" as the maximum number of a particular class repeats divided by the total number of iterations(16). This way the maximum commitment is 1 (particle was assigned to the same class for all 16 iterations) and minimum commitment is 0.375 (as there are only three classes, a particle has to be assigned to some class at least 6 iterations). Below is the histogram of how the data on roughly 140000 particles looked like. I tried throwing out all the particles with "commitment" below 0.5 and re-refining, but it didn't really help in my case.
Hope this helps,
Klim
|