The is all about summer! OK, it’s not generally but the classic explanation of how an association between two observed variables does not prove a causal relationship is that the association between seeing people eating ice cream and wearing short sleeved tops doesn’t prove that ice creams cause short sleeves nor vice versa. The association is probably, assuming that the data was collected over a year or more and that we’re in a country (not Ecuador) where there are seasonal variations in temperature is almost certainly because short sleeves and eating ice creams are each both associated with, probably causally, warmer weather. Here warmer weather, perhaps summer, is the omitted variable.
Details #
It’s an obvious, a trite issue really but in observational quantitative research we can never measure everything: there are always “omitted variables”. The issue is whether they might be plausibly important variables. I to rarely see serious consideration of the issue and I confess that I probably don’t highlight it enough in my own own work. I think we’re scared of acknowledging how incomplete our datasets generally are and how this may influence the implications we draw from the data we have.
Try also #
Causal attribution
Chapters #
Not mentioned in the OMbook.
Online resources #
None.
Dates #
First created 24.viii.24.
