JASP won't run my analysis?

Comments

• František

  April 5

  Hi Lana,

  Sorry to hear that you are having issues running the analysis. Could you maybe check the 'Model' section and simplify the random effect structure? It seems like that you have specified quite a complex GLMM. I would start by removing all the interactions from the random effects and then slowly addding the terms back one by one and see whether the model converges.

  By default, JASP triest to estimate the full model, which in the case of two random effects grouping factors and 3 predictors accounts to 3 way interactions and the corresponding random slopes.

  Cheers,

  Frantisek
• Lana33

  April 5

  I'm presuming this is what you meant? When I try to add any of them back in the analysis wont run again. I was advised to us a standard LMM but even then i still get a warning for model fit is singular so im not really sure at all what's going on.

  

  
  
• František

  April 5

  Yes, this is what I meant. From the output, it seems like your data do not contain enough information to estimate the random effects of both groups and participants. I don't know what does your data structure looks like. I assume you have repeated measures within participants but how does the group random factor play into this?
• Lana33

  April 5 edited April 5

  This is an embed external element. It can be deleted using the delete key or the backspace key. To view the full element, press the preview button below.

  Experiment Data new.csv

  39.08 KB

  This is what my raw data file looks like. So I have 40 rows for each participant (n=30) as they saw 20 words and then I had immediate/delayed recall. The groups are only there as participants were split into groups of six to counterbalance my colour conditions but they saw the same set of words. I've only added group as a random effect as I was advised to use a standard LMM and include both to the random effects box but that wasn't working so I just did the same when I changed to GLMM but wasn't sure if it was necessary or not.
• František

  April 5

  Thanks for sharing the data, this is very helpful! I quickly ran the analyses and it seems like there is no variation between participants, e.g., see the output below (the random intercepts are estimated to have zero variance).

  

  
  

  However, there seems to be some between-groups variance, and more importantly, between word variance.

  

  I'm not sure what exact specification makes the most sense given your research question, but including both subjects and stimuli as random effects is recommended for cognitive psychology experiments. As such, including the words as a random factors would be worthwhile and there seems to be a lot of between-word heterogeneity. On theoretically basis, participants should be included too, however, the data do not seem to provide enough information to estimate the between participant differences (although using the groups only seems to be feasible).

  Cheers,

  Frantisek
• Lana33

  April 5

  Thank you so much for your help i really appreciate it!!!