Welcome!

Sign in with your CogSci, Facebook, Google, or Twitter account.

Or register to create a new account.

We'll use your information only for signing in to this forum.

Supported by

Proof-reading or feedback on JASP-based Bayes in thesis?

Hello everyone,
I used JASP for two of my thesis chapters (doctoral thesis, not MSc). The approach I took is a combined one between Bayesian and frequentist ANOVA in JASP or Friedman's test in R. I described it here: http://forum.cogsci.nl/index.php?p=/discussion/3229/its-good-to-worry-about-mistakes-when-doing-stats#latest

My PhD supervisor is a believing but not a practicing Bayesian, as he puts it. So he won't be able to give me much feedback on the write up.

Which aspects of JASP should I include in my write up?
1) For ANOVA, I was thinking P(M), P(M/data), BF(M), BF10, BF01, %error
2) For t-tests, I was thinking P(M), P(M/data), BF(M), BF10, BF01, %error plus CI

Is that all I need? Should I NOT include any of these? Is there anywhere I could recruit someone to look at maybe one of the chapters so that I can then fix that chapter and the second chapter based on the feedback on the first? Or could I copy-paste some analyses & interpretations in here to get some feedback? I'm really thrilled about JASP but I don't want to interpret it wrong or write it up wrong...

Comments

  • EJEJ Posts: 392

    My advice:
    1. Have an online appendix in which you present the annotated JASP output (and/or put the annotated .jasp file on the OSF).
    2. In the main text, for ANOVA, use "best model on top" and describe the column of BF01s (this is general advice; might be different if you have specific hypotheses in mind).
    3. For t-test, describe the output of the prior&posterior plot with additional info.

    Cheers,
    E.J.

    Thanked by 1eniseg2
  • eniseg2eniseg2 Posts: 18

    Hi EJ,
    sorry to bother you again.
    I did as you suggested so I now have this output, but I'm struggling to wrap my head around it.

    Since we are comparing all other models to the best model (FriendliestMostPopular+JobKnowledge) and the null model is at the bottom, this suggests that the null model is the worst-performing model. Does BF01 here mean support for FriendliestMostPopular+JobKnowledge over the null model, so my data are 516 times more likely to occur under FriendliestMostPopular+JobKnowledge than the null model? I just want to make sure I have the interpretation correct.

    Thank you so much,
    eniseg2

  • EJEJ Posts: 392

    Completely correct!
    E.J.

    Thanked by 1eniseg2
Sign In or Register to comment.