8
$\begingroup$

What do covariate and confounding variables have in common and how do they differ? And what are their specific effects in causal inference? (in statistics and causal inference)

$\endgroup$
1
  • 1
    $\begingroup$ Check out my answer to a nearly identical question (posted after yours) here. Basically confounding variables are a subset of covariates that are along a non-causal pathway from the treatment to the outcome. $\endgroup$
    – Noah
    Commented Jun 10, 2019 at 16:20

1 Answer 1

2
$\begingroup$

This is a complicated question because different fields conceive these types of variables differently, where others make no distinction whatsoever (which is the case for many social sciences fields and subfields).

In statistics, a confound is a variable that is so closely related or associated with another variable that you can’t tell their effects apart. In epidemiology, confounding variables to signify a covariate that is related to both predictors & treatment/exposure.

There are also who focus on the effect of a confounder: "A Confounder is a variable whose presence affects the variables being studied so that the results do not reflect the actual relationship". Pourhoseingholi MA, Baghestani AR, Vahedi M. How to control confounding effects by statistical analysis. Gastroenterol Hepatol Bed Bench. 2012;5(2):79-83.

In practice, however, I have seen quite often the interchangeable use of covariates, confounding, predictor, & controls variables. I also seen the difference in nomenclature representing different theoretical considerations. For example, scholars A may classify variable X as a confounder while scholar B as covariate.

Either way, mathematically, and barring causal inferences, when estimating a linear regression or time-series, there is no difference. That is to say, it’s the same model & you run it the same way in statistical softwares. If you are interested in causal inferences, however, nomenclature start to matter and reflect different models of the phenomenon at hand.

$\endgroup$
2
  • 1
    $\begingroup$ I am looking for meanings in statistics as well as in causal inference. $\endgroup$ Commented Mar 6, 2019 at 4:46
  • 1
    $\begingroup$ I am glad you have changed the text of your question to specify what you are looking for. In this way, possible responders can better evaluate whether their expertise fit with your expectations. $\endgroup$
    – Naaf Faan
    Commented Mar 7, 2019 at 21:33

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.