Subsets from multimodel classifcaiton

2 posts / 0 new
Last post
Arne
Subsets from multimodel classifcaiton

Dear cisTEM team. 

I ran a multimodel refinement and now want to create separate stacks for each volume. I am using the option in the create refinement package for this. However, I do get very different percentages of particles - for example after the multimodel refinement the first map is supposed to contiain 20% of the data - but the number of particles is only 10% after creating the stack.

I guess this is related to the maxmimum likelihood approach?

 

many thanks for your help in advance 

timgrant
Hi Arne,

Hi Arne,

In theory the numbers do not have to be the same.  Each particle is put into every class weighted by it's "occupancy".  These occupancies will add up to 100% over all classes, but need not be 100% for one class and 0% for the rest.  When you extract particles, it takes all particles where the occupancy for that class is highest, e.g. if you have 5 classes where the occupancies for a particle are 21%, 20%, 20%, 20% and 19% - this particle will be assigned to class 1.  

However, in my experience the two numbers do not generally vary by that much (1 or 2 % at most), so the example you give is a little strange.  If you could send me the par files, I can take a more detailed look for you if you would like?

Cheers,

Tim

Log in or register to post comments