SDN members see fewer ads and full resolution images. Join our non-profit community!

power analysis for ZINB regression models

Discussion in 'Psychology [Psy.D. / Ph.D.]' started by psyched2graduate, Aug 24, 2017.

  1. SDN is made possible through member donations, sponsorships, and our volunteers. Learn about SDN's nonprofit mission.
  1. psyched2graduate

    psyched2graduate 2+ Year Member

    54
    21
    May 28, 2015
    Does anyone know if there is a way to run a power analysis for zero-inflated negative binomial regression models? Or general set of guidelines for ensuring adequate power for these models? I found a few articles, but they are largely theoretical in nature. I ran the models in Mplus (which I am new to) and was hoping there was a way to do it in that. Thanks for any help!
     
  2. SDN Members don't see this ad. About the ads.
  3. Ollie123

    Ollie123 10+ Year Member

    4,561
    929
    Feb 19, 2007
    No software that I am aware of can do it. It sounds like you already have the data so post-hoc power analysis is not very meaningful anyways (if that is where you are going with this).

    It wouldn't be elegant, but you could hard code something like this in R. Power analysis isn't rocket science. Generate a few thousand random distributions based on your parameters, apply the chosen analysis to each and count sig p values. That's the essence of it. Elegant solutions would make it so it wouldn't take hours to run, but something like that should get the job done for a one-time-only type thing.
     
  4. psyched2graduate

    psyched2graduate 2+ Year Member

    54
    21
    May 28, 2015
    Thanks, Ollie! I was trying to avoid simulations altogether, but this is what I went with. And yeah, an editor suggested we run this analysis but then later thought that we didn't have the power to do it, hence the post-hoc power analysis.
     
    Last edited: Aug 25, 2017

Share This Page