Again, thank you for this series. I kept up with the tutorial quite easily -a few differences due to versions (I guess) especially for the forest plot but this section was a smooth as the others. I commend you for a great tutorial.
Great question. The main difference I see in my field is that meta-regression typically refers to either continuous predictors or multiple predictors in the same model. In contrast, moderator analysis often refers to a single categorical variable (as demonstrated in this example). The intercept model in this example is what you would often see referred to as meta-regression as it has the intercept present (e.g., one 'level' of the moderator is assigned as the intercept). In my field (educational psychology/educational technology) we often use categorical variables rather than continuous. Due to this, in this example we first assign our variable as a 'factor' so that each is assigned as a discrete level. Second, we run the intercept model (meta-regression w/ our categorical variable) to examine our test of moderators, which tells us if the moderator is significantly different between levels (i.e., is it a significant predictor). The third step in this example (removing the intercept) is to build an ANOVA-like table which is frequently reported in my field for categorical moderators. So that's a long way of saying - in my field, meta-regression would refer to continuous variables or multiple variables in the model with an intercept. In contrast, moderator analysis refers to categorical variables. I hope that helps!
Certainly! Kcomparisons counts the number of rows in your data set for each "level" of the moderator. So in this example in the video, there were 3 effect sizes for grades 6-8, 2 for Not Stated, 18 for post-secondary, etc. I hope that makes sense, if not please let me know.
Thank you!
Again, thank you for this series. I kept up with the tutorial quite easily -a few differences due to versions (I guess) especially for the forest plot but this section was a smooth as the others. I commend you for a great tutorial.
Thanks for the amazing explanation. I have a question, what is the difference between this Moderator Analysis and the "Meta-Regression"?
Great question. The main difference I see in my field is that meta-regression typically refers to either continuous predictors or multiple predictors in the same model. In contrast, moderator analysis often refers to a single categorical variable (as demonstrated in this example). The intercept model in this example is what you would often see referred to as meta-regression as it has the intercept present (e.g., one 'level' of the moderator is assigned as the intercept). In my field (educational psychology/educational technology) we often use categorical variables rather than continuous. Due to this, in this example we first assign our variable as a 'factor' so that each is assigned as a discrete level. Second, we run the intercept model (meta-regression w/ our categorical variable) to examine our test of moderators, which tells us if the moderator is significantly different between levels (i.e., is it a significant predictor). The third step in this example (removing the intercept) is to build an ANOVA-like table which is frequently reported in my field for categorical moderators.
So that's a long way of saying - in my field, meta-regression would refer to continuous variables or multiple variables in the model with an intercept. In contrast, moderator analysis refers to categorical variables.
I hope that helps!
@@LearnMetaAnalysis Thanks! this was very helpful
Could you kindly explain what the 'kcomparison' column is supposed to communicate?
Certainly! Kcomparisons counts the number of rows in your data set for each "level" of the moderator. So in this example in the video, there were 3 effect sizes for grades 6-8, 2 for Not Stated, 18 for post-secondary, etc. I hope that makes sense, if not please let me know.
@@LearnMetaAnalysis Yes please it does. Thank you, thank you!