Posted by Ray Poynter, 3 June 2019
What about Partial Correlation?
Ray recently released the slides and recording of an introduction to using correlation for finding stories in the data – access them here.
I’ve received a query about Partial Correlations and causality. Here is a copy of my reply.
Partial correlation does not require or indicate causality, but researchers will often assume causality (for example from some other modeling or logic) when using it.
Partial correlation looks at the correlation between two variables when controlling (i.e. eliminating) the effect of one or more other variables.
For example, a researcher might want to look at the correlation between the price and sales of ice cream, when controlling for temperature. In doing this the researcher might assume that sales is the dependent variable, that the price is the independent variable, and that temperature is the variable whose contribution she/he wishes to remove.
There is a nice description of partial correlation and the assumptions that underpin it on the Laerd Statisitcs site.