Initial values for estimates during bootstrap

[Jiaqi Yang]

Dear Laurent and all,

I am using FastSimCoal v28 to jointly estimate 15 parameters for a model with 4 populations: fsc28 -t file.tpl -e file.est 0L 40 -n 1000000 -M -d --multiSFS -s0

I first ran using the full dataset with 200 runs and chose the set of parameters with the highest likelihoods as the best parameters.  

For each bootstrap replicate (the full dataset was divided into 500 blocks for bootstrap),  I used the best parameters above as the initial parameter values with-- initValues and performed 20 independent runs for each bootstrap replicate.

It turned out that  5 of the 15 parameters were almost EXACTLY the same across bootstrap replicates and across runs of each bootstrap.  While the values make sense, the distribution of those 5 parameters across boostraps/runs look weird to me. I was expecting some variation at least.

Does anyone have thoughts on this? I would appreciate any ideas.

Thanks!