Dear all,
I am running a series of PLS analyses on distinct ERP data.
What is the best way to set the pos/neg threshold for the bootstrap results? is it preferable to keep the default value for each analysis, or to set a unique value for all of the analyses? how is the default value obtained?
Many thanks,
Giorgia
Dear all,
I am running a series of PLS analyses on distinct ERP data.
What is the best way to set the pos/neg threshold for the bootstrap results? is it preferable to keep the default value for each analysis, or to set a unique value for all of the analyses? how is the default value obtained?
Many thanks,
Giorgia
Hi Giorgia,
There is no strict rule on where to set the threshold. You should be mindful that these ratio are not meant for null hypothesis testing - that is the done with permutation. The ratio are meant to identify the timepoints/channels with the most robust contribution to the overall effect - it helps you describe the effect. Treat them as confidence intervals rather than something akin to a z-test.
You do not need to set the same threshold for postive and negative values since there is no assumption of symmetry in the distribtion. The default value is determine by calculating the 2.5% and 97.5% value of the empirical bootstrap distribution so will differ depending on the LV and data set. "Rough" rules of thumb: threshold of 2.3 is about 95% confidence interval and 3.09 is 99%.
Nancy and I wrote a review a few years back that has a section on the bootstrap ratios and how to think about this
https://www.dropbox.com/s/pwszup8m0f92us3/pls_review.pdf
Dear Randy,
thank you very much for your reply!
Giorgia
Baycrest is an academic health sciences centre fully affiliated with the University of Toronto
Privacy Statement - Disclaimer - © 1989-2024 BAYCREST HEALTH SCIENCE. ALL RIGHTS RESERVED