As best as I can tell, linear regressions are used throughout. This is not an appropriate method for when the dependent variable is binary (e.g. ‘mentally ill or not’. Logistic regression should be used instead—and may find some further associations, given linear methods will intuitively be underpowered.
As it happens, running a logistic instead of a linear regression for ‘mentally ill or not’ ~ involvement in EA doesn’t change the results. I’ll run the others at some point.
Thanks for this. A statistical note:
As best as I can tell, linear regressions are used throughout. This is not an appropriate method for when the dependent variable is binary (e.g. ‘mentally ill or not’. Logistic regression should be used instead—and may find some further associations, given linear methods will intuitively be underpowered.
Thanks!
As it happens, running a logistic instead of a linear regression for ‘mentally ill or not’ ~ involvement in EA doesn’t change the results. I’ll run the others at some point.
Try with
family = "binomial"