Dear Stata Professionals,

i have a question on a meta-analysis problem.

I have 38 studies on which I want to do a meta analysis.
Out of the 38 studies I built 2 Variables and computed the effect sizes from these 2 Variables called A and B in their relation to the Dependent Variable C.

The hypothesis sound like:
H1: A has a positive Effect on C.
H2: B has a positive Effect on C.
H3: The effect of A on C is stronger than the effect of B on C.

I want to test whether the effect of A on C is stronger than the effect of B on C.

Array

What I already did was to test the effect of A on C and A on B with the 'meanes' command:

generate sd_a_on_c = sqrt(((1-( r_a_on_c^2))^2) / ( n_size -1)), after( r_a_on_c )
generate sd_b_on_c = sqrt(((1-( r_b_on_c^2))^2) / ( n_size -1)), after( r_b_on_c )

generate z_a_on_c = 0.5 * ln((1+ r_a_on_c )/(1- r_a_on_c )), after( sd_a_on_c )
generate z_b_on_c = 0.5 * ln((1+ r_b_on_c )/(1- r_b_on_c )), after( sd_b_on_c )


generate se_z_a_on_c = sqrt((1 / (n_size - 3))) if r_a_on_c != ., after( z_a_on_c )
generate se_z_b_on_c = sqrt((1 / (n_size - 3))) if r_b_on_c != ., after( z_b_on_c )

generate invweight = n_size - 3 if r_a_on_c != . | r_b_on_c != .

*** Meta-Analysis -meanes- command
meanes z_a_on_c [w=invweight], print(ivzr)
meanes z_b_on_c [w=invweight], print(ivzr)


My output looked like this:

Array

A --> C is 0.046 and 95%CI excludes zero
B --> C is 0.006 and 95%CI includes zero

Now I want to compare A --> C with B --> C and see whether the effect of A is stronger than B - eventhough B-->C is not signficant - because this would be the hypothestis in the paper.



Can you help me and tell me which command or user written programm i should use?

I also provided an excel sheet with an example on how the data looks like.


Thank you so much in advance!!


Kind regards
Svenja