Bootstrapping Network
Hi folks,
I'm having a hard time figuring out how the bootstrapping works in JASP. I'm using "mgm" estimator, "EBIC" criterion and "AND" as well as "OR" rule. After conducting bootstrapping (I need non-parametric, but the effect is the same for all types) the results stay exactly the same for sparsity, centrality measures, clustering measures and weights matrix, no matter if I use 10 or 1000 bootstraps. Also the network graph is unchanged. I do get the bootstrap network summary and two edge stability graphs though.
For CV criterion bootstrapping results in changes in all measures and graphs.
I'd really appreciate your support as I couldn't find solutions so far.
Best regards, Chris
Comments
Hi Chris,
Hmm that's strange. I'll contact our network expert.
Cheers,
E.J.
Hi Chris,
The bootstrapping is only used to assess the stability and accuracy of the results, not to estimate the networks themselves. This is why the results are always the same. If you use cross-validation as estimator, the results differ a bit each time you run an analysis due to a different random seed used in selecting the cross-validation folds.
Best,
Sacha
Thanks guys, really appreciate your answers, helps a lot!
Best, Chris