Accelerated Bootstrap - mediation

[danony]

Howdy! :tup:

Can anyone explain accelerated bootstrap to me? I've got it reported in a research report and I can't really figure it out. Basically a couple of covariates were significant between groups so they ran a mediation analysis on them to see if they could account for the pattern of results. But they used this accelerated bootstrap procedure which I can't even find anything on in Tabachnik and Fidel!

Cheers,

[spunky]

it's a type of estimate that you use when the boostrap distribution has a bias, non constant variance, etc... in simpler terms, if the boostrap distribution is not symmetric chances are the endopints of the confidence intervals may be a little off and you'd probably be better off considering the accelerated boostrap...

for a better intro to the bootstrap method: http://www.wiley.com/legacy/wileychi/eoenv/pdf/Vab028-.pdf

(ps- just between you and i, T & B is a pretty limited to only cover the essentials, they leave out a lot of the stuff you need to understand more advanced techniques... ;)

[danony]

Thanks spunky, that is a helpful reference - I've saved it for future reference too! I also found preacher and hayes 2004 on mediation and that's helped me 'get it' a bit more too. :)