Non-independent sub-group analysis in meta-analysis

[Buse Aras]

Hello,

I am conducting a meta-analysis examining the effect of childhood maltreatment on emotional reactivity. Additionally, I aim to compare the effects of different subtypes of childhood maltreatment on emotional reactivity.

There are five subtypes of childhood maltreatment: physical abuse, sexual abuse, emotional abuse, physical neglect, and emotional neglect. Because studies often report multiple correlation coefficients from the same participants (e.g., one study may report correlations between three different maltreatment subtypes and emotional reactivity), my dataset includes non-independent effect sizes. To address this, I am using a non-independent subgroup analysis within the study model.

I am able to perform the overall analysis to examine the general effect of childhood maltreatment on emotional reactivity. However, I am encountering an issue when trying to compare the five maltreatment subtypes. Specifically, Stata selects one maltreatment subtype as a reference category and only displays the results for the remaining four subtypes. I need to see the effect sizes for all five maltreatment subtypes and compare their effects directly, but I am unsure how to modify my analysis to achieve this.

I have attached:

• An example from my dataset for clarification.
• A screenshot of my Stata output showing the issue.

I would greatly appreciate any guidance or suggestions on how to properly conduct the subgroup analysis to include all five maltreatment subtypes.

Thank you in advance for your help!

[Andrew Musau]

To get the adjusted predictions, use margins following your mixed command.

mixed fisher_z i.maltreatment || Study_ID:, variance
margins maltreatment