How to compute the 95% confidence intervals?

[Ruby Rao]

Dear Laurent and fastsimcoal users,

I have already completed bootstrapping parameters for my chosen best model, and I obtained 100 boottrap replicates, each with 30 runs ,  resulting in 3000 bootstrap values per parameter of interest. 

 I am now confused about whether I should compute the 95% confidence intervals (95%CI) based on the 3000 bootstrap values, or only on the  100 best estimates  (selected from each set of 30 runs base on MaxEstLhood value) ?

Furthermore, what's the exact computational  formula for the 95% CI? Different articles provide different explanations,  and I am still unsure about the correct approach.

Could you please help me with this issue?

Many Thanks,

Ruby