Hallo,

I'm wondering if there is a way to compare the magnitude of two different associations, each with the same predictor variable (and covariates) but different dependent variables. To give a (fictional) example:

Imagine that I have some survey data in which participants were asked how much they liked cats (scale of 1 to 10), how much chocolate they ate (grams per month), how much orange juice they drank (litres per month), and some covariates (age and gender).

First I want to see if liking cats is cross-sectionally associated with eating chocolate. To do that, my plan is to run a multiple linear regression including liking cats, eating chocolate, and the covariates. [Apologies if I've already gone wrong at this step!]

I also want to see if liking cats is associated with drinking orange juice. To do that, my plan is to run a regression including liking cats, drinking orange juice, and age and gender. So the same model, except this time with orange juice instead of chocolate as the dependent variable.

Finally (the focus of my question), I want to see if liking cats is *more strongly* associated with eating chocolate than it is with drinking orange juice.

Is there a way to do that?

My idea was to compare standardised regression coefficients for each (cats*chocolate and cats*orange juice) - however I'm not sure if this is valid at all (ie to note that one is larger than the other), and even if it is, I'm not sure how to test for a statistical difference between the two.

This seems to be easily done with correlation coefficients, but I can't find a clear explanation as to whether it's feasible, and if so how to do it, with regression coefficients (or some other way to do it when you want to include covariates). Is it doable? Or is there another analysis to find out if the effect there is a difference in the magnitude of the associations?

Apologies if this is a stupid question. Many thanks in advance for any help anyone can give.

I'm wondering if there is a way to compare the magnitude of two different associations, each with the same predictor variable (and covariates) but different dependent variables. To give a (fictional) example:

Imagine that I have some survey data in which participants were asked how much they liked cats (scale of 1 to 10), how much chocolate they ate (grams per month), how much orange juice they drank (litres per month), and some covariates (age and gender).

First I want to see if liking cats is cross-sectionally associated with eating chocolate. To do that, my plan is to run a multiple linear regression including liking cats, eating chocolate, and the covariates. [Apologies if I've already gone wrong at this step!]

I also want to see if liking cats is associated with drinking orange juice. To do that, my plan is to run a regression including liking cats, drinking orange juice, and age and gender. So the same model, except this time with orange juice instead of chocolate as the dependent variable.

Finally (the focus of my question), I want to see if liking cats is *more strongly* associated with eating chocolate than it is with drinking orange juice.

Is there a way to do that?

My idea was to compare standardised regression coefficients for each (cats*chocolate and cats*orange juice) - however I'm not sure if this is valid at all (ie to note that one is larger than the other), and even if it is, I'm not sure how to test for a statistical difference between the two.

This seems to be easily done with correlation coefficients, but I can't find a clear explanation as to whether it's feasible, and if so how to do it, with regression coefficients (or some other way to do it when you want to include covariates). Is it doable? Or is there another analysis to find out if the effect there is a difference in the magnitude of the associations?

Apologies if this is a stupid question. Many thanks in advance for any help anyone can give.

Last edited: