Incorrect mean-difference calculation
I'm running a rather large, complicated ANOVA, and I notice that the mean-differences for the post hoc test are not being calculated correctly. I double checked by creating a new CVS file, and a new JASP file from scratch and ran in to the same miscalculation both times. Attached is a screenshot showing the error. You can see the two values 30.011 and 29.310, which the difference is 0.701. On the right you can that I assigned it to the correct cell. The difference is from the between-subject factor "compare to".
I double checked all the ns, means, and STDs with an excel spreadsheet and it was the same data that was loaded into JASP. However, all the mean differences between JASP's output, and what I calculated on Excel are different. I've spent the last three hours checking the inputs, confirming the columns on my excel spreadsheet are the correct ones that went into JASP. But you can see here, plain as day, that the two mean differences are not correct in the t-test result.
Not really sure what to do here. This is a much more complicated ANOVA than I've typically run. Usually I have 2-3 independent variables, and I manually calculate part of the results in Excel or MATLAB to make sure JASP's output is correct, which they always are.
I'm also including my CSV spreadsheet
Comments
I think it's because those are so-called adjusted means, i.e. means adjusted for the effects of the other factors included in your ANOVA. What happens if you test with a simple one-way ANOVA (only your factor)? In that case the means probably match the sample means?
Thanks. I did end up looking further into this and the way JASP weighs means differs from how I assumed it would. After taking that into consideration, I replicated this on my end.
@samamarc ,
Great to hear this is clarified now. A blogpost I wrote a while ago might illuminate the issue some more:
Cheers
Johnny