Meta-Analysis with R (Episode 44): Overfitting, Multicollinearity and Model Building

Why do meta-regressions so easily see signals in noise, and why do highly correlated predictors make them unstable? The video shows overfitting with a schematic picture, gives four reasons why meta-regression is especially at risk and four safeguards, explains multicollinearity with a small simulated example, and compares ways to build the model: forced entry, hierarchical, step-wise and multimodel inference. All pictures are schematic, with made-up numbers. Video 44 of the series, fifth of eight on chapter 8 of the book 'Doing Meta-Analysis with R'. It follows the free online textbook by Harrer, Cuijpers, Furukawa and Ebert (https://doing-meta.guide); the videos give a first overview, the book goes into depth. This video series accompanies courses at the Department of Psychological Research Methods, Trier University: http://bosnjak.uni-trier.de Next: Multiple Meta-Regression in R: rma and anova    • Meta-Analysis with R (Episode 45): Interac...   Playlist of the course:    • Doing Meta-Analysis with R: A Hands-On Vid...