Calculation of point estimates from bootstrapping

[jonatha...@gmail.com]

I'm calculating point estimates and standard errors using se = bootstrap in lavaan. 

How does lavaan calculate point estimates and standard errors from the boostrapped sampling distribution? It appears as though standard errors equal the standard deviation of the boostrapped sample, but the point estimates do not appear to equal the mean (at least not for variances). Which statistic used to calculate point estimates?

Thank you!

-Jon Helm

[Terrence Jorgensen]

How does lavaan calculate point estimates and standard errors from the boostrapped sampling distribution? It appears as though standard errors equal the standard deviation of the boostrapped sample, but the point estimates do not appear to equal the mean (at least not for variances). Which statistic used to calculate point estimates?

se = "bootstrap" only requests bootstrap standard errors.  Since only SEs and test statistics seem to be substantially biased by nonnormality (according to simulations I am aware of, happy to be corrected), there is no need to correct that, so the default output for point estimates is the ML (or LS) estimate).  But if you use bootstrapLavaan(), you can of course report whatever central tendency you prefer.  I think that function still does not provide the option of bias-correction with acceleration (I forget the reason), so if you want that you could instead use the boot package and write a custom function to extract whatever information you are interested in.

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen