Rather a pompous but lovely term for a very common mistake. It means assuming that something from an analysis of group data applies for an individual (whether in the dataset or not).
Details #
A typical example is to assume that because in an analysis of a dataset from a service shows a strong relationship between baseline score on a measure and immediate rate of improvement on that measure, measured say sessionally or weekly that it can be assumed that your new client with a very high baseline score will show rapid improvement. There is no guarantee of that, yes, over that dataset and perhaps much more widely (it is a not uncommon finding) that relationship holds but because relationships in our field are stochastic, statistical, variable (the reason why we use statistics to analyse them) the ecological fallacy is a fallacy.
Note that the only real exception to this is when you have a perfect 1:1 relationship in your data. When I was just starting in research the lovely statistician in the department (of psychiatry as then was) told me that on an occasion before a visiting researcher had been seen and heard running down the corridor shouting “I’ve got a p < .001” … only to have it pointed out that this was the correlation between date of birth and age (presumably calculated to the day). There you have an aggregate relationship that will apply to everyone. Mostly we don’t have that!
Try also #
- I am not sure there is any more to be said!
Chapters #
Not covered in the OMbook though with hindsight we probably should have covered it.
Online resources #
There is more on it and other statistical fallacies at https://en.wikipedia.org/wiki/Ecological_fallacy.
Dates #
First created 22.ix.26.
