Comparing results by platform

[dnmacdon]

@rmarkello As a part of this project, I may end up rerunning some linear model and effect size calculations in Python that I previously ran in R. I would expect to see exactly the same results, but if I don't, any suggestions on metrics that I could use to evaluate how similar the results are?

[rmarkello]

Hi @dnmacdon ! As you said, you should be able to get exactly the same results for most simple linear models + effect size calculations. If you don't I think I would recommend spending some time understanding why (and resolving those issues) rather than trying to quantify the differences.

Things to watch out for are e.g., differences in calculation of population versus sample standard deviation (Python generally defaults to population whereas other languages will often use sample—I'm not sure about R). If you get differences and aren't sure where they're coming from, let me know and I'd be happy to dig into it with you!