search

elmacartney / EE_stress_MA

MIT License
0 stars 0 forks source link

Function for transforming proportions #20

Closed elmacartney closed 3 years ago

elmacartney commented 3 years ago

@itchyshin there is now a column in the data called "Response_percent" where yes = percent.

Can you please add the function that transforms these and add how this was done in the methods?

Thanks!

itchyshin commented 3 years ago

@elmacartney - I have done it

A new function is in Functions_cleanedup.R

Also, I changed RMD file so we now get the correct effect sizes.

I did find another mistakes (my original ones which I got you to check) so it it probably best we go through it together and I will also ask how you actually check the values last time.

Please check these and see whether it makes sense.

Also, actually pairwise analyses may become significant (this is I had a mistake) - although generally - this has less power than the main effect but each of these is probably more powerful than interaction (interaction actually turns out least powerful - as estimating this a bit more complicated - they say we need 16 times more sample size to estimate interaction correctly - a perfect for meta-analysis - please bring this up next time we need - I want you to understand this). Maybe we can quickly catch up once you rerun the analyses.

elmacartney commented 3 years ago

@itchyshin just letting you know that I've rerun everything and results are pretty much consistent, including the interaction. Like you said, pairwise comparisons are now significant (apart from ES/EC) and heterogeneity has gone up substantially.

We can discuss more on Monday.

itchyshin commented 3 years ago

@elmacartney - thanks for an update - I will update a method section with regard to power a bit soon!