I am doing sem
with lavaan
in R, and I found that even I don't input the bootstrap parameters, the output of sem
will give me the standard error by default.
I am wondering what is the difference between the default output standard error and the bootstrapped standard error?
Best Answer
It depends on the type of estimator. The default without
ordered=
outcomes is maximum likelihood, in which case the second derivative of the likelihood function (Hessian matrix) is used to obtain the SEs analytically. There are different flavors depending on (in)complete data, or robust corrections for nonnormality. This recent paper provides and nice overview of the options:https://doi.org/10.1080/10705511.2021.1877548
A bootstrap SE is simply the SD of a statistic in the empirical bootstrap distribution.