-1

I understand that correlation can tell us the association (strength, direction) between two continuous variables. I also understand that Regression analysis can predict the behavior of the dependent variable based on the independent variable.

What I am not clear is which statistical method should we use to determine if variable A has an effect on variable B, or vice-versa?

In my study, I am interested to know if usability (independent var) has an influence on affective responses (dependent var), and also whether affective state (independent var) has an effect/influence on perceived usability (dependent variable).

Vyas
  • 309
  • 1
  • 4
  • 11
  • 1
    You've tagged 'causality'. Have you taken a look at the questions in that tag? https://stats.stackexchange.com/questions/tagged/causality?sort=votes&pageSize=50 – mkt Apr 04 '18 at 13:43

1 Answers1

1

That's the subject Pearl's Causality. If you can do experiments, i.e., you can control the independent variable while holding everything else constant and observe the dependent variable, then correlation proves causation. When you can't do this, there are other techniques, which I tried to summarize here.

Neil G
  • 13,633
  • 3
  • 41
  • 84
  • +1 because this post will be helpful to the OP and not wrong in general but think your answer can mislead some readers to think that Pearl discovered/established "Causality" as a theoretical notion. The idea of "[causation and manipulability](https://plato.stanford.edu/entries/causation-mani/)" was well established before his time, "[probabilistic causation](https://plato.stanford.edu/entries/causation-mani/)" was also clearly touched upon multiple times before too. JP did/does have huge influence in the field but did not invented it. ("*GO SGS!*" I kid.. I kid... :) ) – usεr11852 Apr 08 '18 at 00:02
  • The [Primer book](http://bayes.cs.ucla.edu/PRIMER/) might be easier intro to this. – dimitriy Apr 11 '18 at 18:06