[R-meta] General guidance, control vs. moderator variables in meta-regression

[Farzad Keyhan]

Dear Meta-Analysis Colleagues,

I'm meta-analyzing a group of studies to determine the extent to which
a family of instructional strategies (X) can cause an improvement in
students' writing outcomes (W).

Since these studies are quasi-experiments, the study groups (X) have
not been randomly formed, thus, I want to control for potential
confounds in these studies.

Although I have seven variables in mind that can affect X --> W as my
main causal query, is there a general guideline/way of thinking to
determine which one of these are moderators vs. control (and not
interpreted)?

Any help/insight is appreciated,
Fred

Here are my seven variables:

1- How many times the instructional strategies were provided to
students-->continuous
2- What type of outcome[s] the strategies targeted-->categorical
3- How many distinct outcome[s] the strategies targeted-->continuous
4- How many times the effect of strategies were measured (1 post-test,
2 post-tests etc.).
5- What were students' writing proficiency level (beginning,
intermediate, advanced)-->categorical
6- Where was the study done at (high school, university,
elsewhere)-->categorical
7- What was the genre in which the writing outcomes was elicited
(narrative, argumentative)-->categorical