Hi, In the Phoenix help file, Stderr is defined as standard error. However, when I export the rawout.csv of a bootstrap run, and perform the summary stats myself, it seems that the values reported under the Stderr are actually the standard deviations and not the standard errors. Could you comment on the discrepancy? Similarly, the BootOmegaStderror worksheet seems to contain the standard deviations, and not the standard errors. Also, what does stderr from the Theta worksheet represent? I always thought that it is also standard error, and therefore, the values in the CV% column represent the %RSE. Now I am confused…are these RSEs, or CVs? 
Thanks, Dora
Hi Dora, Given a bootstrap distribution you can compute the following: replicate 1 CL value 1 . . replicate 1000 CL value 1000 The standard deviation of the above vector is an esimate of the standard error on CL this is standard from bootstrap theory. Better is to report the say 95 % confidence interval on CL so you take the 2.5 % and 97.5 % percentiles of the above 1000 values We typically do the bootstrap to visualize the bootstrap distiubtion and look at the histograms and also to derive a 95 % CI. We are not that interested in the CV % or in the standard error per se. In the standard Theta worksheet the standard error is taken from the square root of the diagonals of the variance covaraince matrix this is standard from asymptotic and maximum likelihood theory. So I confirm that the CV% is the % RSE. Note that when we have a parameter estimate and its standard error we assume normality and then we construct 95 % CI by saying it is : estiamte ± 1.95 *SE this is making a good amount of assumption while in the boostrap you take the whole dsitribution compute the quantiles with less assumption made about nornality or perfect symmetry of the CI Samer
Dear Dora Nothing to be confused about. When you bootstrap and fit each bootstrap, you get information about the uncertainty of the model parameters. The standard deviation obtained by running descriptive statistics on the model parameters is an estimate of the uncertainty in that model parameters, uncertainty quantified by standard error. In fact this standard error should be similar to the one obtained from the reference data set from which the bootstrap data sets were generated. Best Serge