0
$\begingroup$

Beginner question sorry - I'm a coder and need stats advice.

I have a dataset broken down by local area, with columns for the proportion of owners who are French, the proportion of owners who grow artichokes, and the mean house price:

   area,proportion_owners_french,proportion_growing_artichokes,mean_house_price
   1,0.4,0.2,352512
   2,0.34,0.1,276241
   ....

I want to see if there's a correlation between the proportion of owners who are French, and the proportion who grow artichokes.

But I also want to remove the effect of house prices, because it's possible that house price is a confounder - people who live in expensive houses could independently be both more likely to be French and more likely to grow artichokes!

What statistical technique should I be using to do this?

Thank you for your advice.

$\endgroup$

1 Answer 1

2
$\begingroup$

Two ideas that come to mind are regression and partial correlations. The latter is more precisely an answer to your question,

Regression assumes that one variable is dependent and the others are independent. Here, it makes some sense to say that growing artichokes is dependent and being French is independent. Note that this isn't really a matter of statistics but of reasoning. You might grow artichokes "because" you speak French but it's a little silly to say you speak French "because" you grow artichokes! (Note that I put "because" in quotes. I'm not implying causation, but English, AFAIK, doesn't have a good word for this). (Also, it's possible that someone actually does learn French because they love artichokes, but it doesn't seem very likely).

Correlation treats all variables the same. Wikipedia has an article with details, including methods of computation, and an implementation in R.

$\endgroup$
2
  • $\begingroup$ Thank you! Yes sorry, I'm assuming a dependent variable - I want to examine if being French makes you more likely to grow artichokes, regardless of how expensive your house is. If I do a regression, how do I factor out house prices? $\endgroup$
    – Richard
    Feb 14 at 11:49
  • $\begingroup$ You would add them as a covariate. $\endgroup$
    – Peter Flom
    Feb 14 at 12:15

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.