There are a lot of contextual factors that can influence how to deal with this. But, at a minimum, I would transparently report what I did and what I found in the manuscript and how i dealt with it in the manuscript. E.g., I would state how I checked for outliers and how many were significant, and how i checked for influence and how many were significant, then what decision I made based on these results and why. I'm sorry I do not want to provide instructions on how to specifically deal with it outside of that because every field has different norms and 'stakes' involved with your decision. E.g., if you were to change an ES because it is an outlier, it could change your results.
@@LearnMetaAnalysis I took this approach eventually: After doing influence: If any test is significant I would report it's results as it is and say this is potentially and outlier. Then I would do a Leave one out and add the Forest plot as a supplementary for extra documentation and information. (I report if the significance of the effect changes). I couldn't find "norms" about this because I can find all sorts of opinions and types of MAs done.
Wow, this video was so practical!
I am going to watch this several times!
thank you
If every outlier I found (student) non were significant on cook's distance. what would you do here?
There are a lot of contextual factors that can influence how to deal with this. But, at a minimum, I would transparently report what I did and what I found in the manuscript and how i dealt with it in the manuscript. E.g., I would state how I checked for outliers and how many were significant, and how i checked for influence and how many were significant, then what decision I made based on these results and why.
I'm sorry I do not want to provide instructions on how to specifically deal with it outside of that because every field has different norms and 'stakes' involved with your decision. E.g., if you were to change an ES because it is an outlier, it could change your results.
@@LearnMetaAnalysis I took this approach eventually:
After doing influence:
If any test is significant I would report it's results as it is and say this is potentially and outlier.
Then I would do a Leave one out and add the Forest plot as a supplementary for extra documentation and information. (I report if the significance of the effect changes).
I couldn't find "norms" about this because I can find all sorts of opinions and types of MAs done.