|
|
Which kind of bootstrapping? |
|
Message/Author |
|
|
I was curious if there was some guidance on which boostrapping option to use in different circumstances for cinterval. Let's assume I am frequentist and thus am choosing between symmetric, bootstrap, and bcbootstrap. I know bcboostrap is bias-corrected .... which I have often used for testing indirect effects (as those sampling distributions are not normal). I am intrested in other things, such as regression slopes and differences in slopes in multigroup contexts (e.g., conditions of an experiment). Would the "standard" option be (bootstrap) as that does not force it into any specific format (e.g., symmetric, bias-corrected, etc.)? |
|
|
See writings by David MacKinnon. I think he recommends BCBOOTSTRAP. Symmetric confidence intervals are typically correct for most parameters. It is parameters that are products or odds ratios when bootstrap may be better. Note that Bayes gives non-symmetric confidence intervals without the need to bootstrap. |
|
|
Thanks! That was exactly what I needed. |
|
Back to top |
|
|