1
$\begingroup$

I have a manufacturing problem with input variable I, intermediate additive A and output O. I have observational data of these variables.

Both I and A can impact O, to some extent. Moreover, A is partly determined by I in the data I have. That is, based on the input level, additive levels are partly determined (there's a trend), but not fully.

O = f(I, A) + error_f, A = g(I) + error_g.

My goal is to find the effect of A on O, for different levels of I.

How to frame it in terms of causal analysis? I tried to read about mediators and moderators, but they do not seem to really fit my problem. Also, it's impossible to perform randomized experiments due to practical reasons, so I am dependent on observational data.

$\endgroup$
2
  • $\begingroup$ What kind of scales have your variables? $\endgroup$
    – POC
    Mar 5, 2021 at 18:15
  • $\begingroup$ they are continuous, positive $\endgroup$
    – learning
    Mar 5, 2021 at 18:23

1 Answer 1

1
$\begingroup$

If you want to see the effect of A on O at different levels of I then moderation analysis is what you are looking for. The model would be $O=A+I+A*I+e_O$.

$\endgroup$
2
  • $\begingroup$ Thanks. How would this regression equation differ from that for mediation analysis? $\endgroup$
    – learning
    Mar 7, 2021 at 14:45
  • $\begingroup$ A lot actually. see stats.stackexchange.com/a/309386/102655 there is a lot of documentation on this topic. $\endgroup$
    – POC
    Mar 7, 2021 at 17:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.