Analysis of a mixture of two normals distribution with random effects

[Ruhi]

Hi 

I am stuck with analysis of a data which seems a mixture of two normals distributions.

Experimental design: So we gatherad data on pelvic torsion on subjects, under 5 diffrent conditions, 3 trials of each condition and  each condition will generate 12 images, so 12 objective records for each person at each trial and condition level. A person's pelvis could have torsion in positive direction  and others could just have it in negative direction, generating a bi-modal distribution shown below. 

I want to model the mean torsion for each different condition. My resaerch into mixture models suggested that I could use proc FMM in SAS, but it cant handle correlated observations and random effects. As I have random trials and images with in each trial, I am not sure how should I analyse this data. 

Any help is very much appreciated? 

Thanks

[PGStats]

What we are looking at is the distribution of the data (effect + error). What you need to model is the distribution of the errors. What does that distribution look like after you substract the mean for each condition?

[Ruhi]

Here is the distribution after subtracting the means from each condition. It is still multi-modal.