1
$\begingroup$

I want to control for the effects of confounds in an observational study (using self-report questionnaires), however, I do not want to imply a causal relationship between my two constructs. Therefore, I am unsure if I can conduct a regression without inferring causality, and so I do not know how to control for confounds if I am only running a Pearson's r.

Let's say I'm looking at emotional awareness (construct 1) and empathy (construct 2). It could be assumed that greater emotional awareness may correspond to increased levels of empathy; however, this is not necessarily supported in the literature.

I want to analyse the strength (and presence) of a relationship using Pearson's r, but I also want to then control for the possible (and likely) effect of sex and age using a regression. Trouble is, both emotional awareness and empathy are independent variables...can I treat empathy as a dependent variable without impacting statistical validity?

Thankyou for any responses! I'm new to statistics.

$\endgroup$

1 Answer 1

3
$\begingroup$

You can run a regression analysis with empathy as dependent and emotional awareness, age and sex as independent variables. Regression does not assume causality between the independent and dependent variable (although independent variables are often referred to as "predictors" and we also say "X predicts Y", but this is mostly just shorthand for describing the model). If you find a significant relationship in your regression analysis, you have evidence that emotional awareness and empathy are related while controlling for age and sex, not that emotional awareness causes or causally affects empathy.

The possibility of inferring causality comes from your research design, not from the test you use. If you have a well-controlled, randomized experimental design, you can find evidence of causality using one of several different statistical tests. From a correlational design, you can't infer causality regardless of what test you use.

$\endgroup$
1
  • 1
    $\begingroup$ Great, thank-you so much for your help. This explains exactly what I needed to know! $\endgroup$
    – Liam C
    Jun 21, 2023 at 18:31

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.