How to do Three Level Meta-Analysis in R: Part 2 - Run meta-analysis, heterogeneity, outliers etc

[@alirezahekmati7632]

Hello! After finishing the other playlists, I will watch the three-level Analysis playlist. Thank you for the work you are doing. Keep up the good work!

[@georgebashour4333]

Great video ❤ Outliers and influences are really the most trouble I had in getting into Meta-analysis.. So many different opinions and my OCD hates this high level of uncertainty 😂😂 I went with this explanation of my Data in my last conventional Meta-analysis and did not remove any study: 1- Did the Influence function from metafor on all outcomes (only reported the significant tests which were mainly studentized res and cook's distance) 2- Did a leave one out (LOO) if there was any outlier study And reported as following in every outcome (same writing aswell) :(still first draft) "An examination of the studentized residuals revealed that one study (Pandit 2024) had a value larger than 2.3940 with a cook’s distance of 1.0433 and maybe a potential outlier in the context of this model. However, the study was not found to be notably different from others, so the effect was not removed or modified. LOO revealed that the overall effect size would become significant if the outlier was held out (Supplementary Fig)."

[@LearnMetaAnalysis]

The uncertainty really drives me crazy as well! That's why I now view outliers/influence similar to how I view publication bias: contextual information that helps me understand my analysis. That has helped my brain a lot since i switched over to that mindset. It's particularly challenging with 3 level models because as far as I know (i could be wrong) there are not 'known' values for what is an outlier or influential study specifically in the context of 3 level MA; what I presented in the series is the same as conventional MA. With publication bias we know conventional publication bias tests don't work well but there isn't really a 'perfect' solution I know of. The one paper that i know of that investigated this suggested a variant of Egger's regression for some circumstances: psycnet.apa.org/fulltext/2020-52356-001.html

[@FC-os1mm]

Hello, thank you very much as always. May I ask if there are any videos on CHE and RVE? You mentioned it will be in future videos. ^^ No pressure, I just love your explanations. Thank you very much

[@LearnMetaAnalysis]

My work schedule has been very busy so I have not been able to create it yet. If you want the code, you can use Simple Meta-Analysis to run a 3level MA with CHE RVE and it will show you the Rscript for each analysis: noahschroeder.shinyapps.io/SimpleMeta-Analysis/ If you prefer a text format, it is described in Harrer et al.'s open book: bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/multilevel-ma.html#rve however their example uses impute_covariance_matrix from clubSandwich which has been deprecated. That is why SMA uses vcalc from metafor instead. Unfortunately I don't think i'll have time to make a video about it for at least a few weeks as I have a few notable deadlines approaching.