Question by my sign: Can I use a logistic regression?
Can I use logistic regression for a pre and post design?
I am using regression because there are control variables and I was advised to use logistic regression because the outcome distribution is skewed so we transformed to “yes/no”.
I have never used this method before so I did some Web research and found a source that said that one of the assumptions of logistic regression is “independence of errors – assumes a between subjects design” – I have a within subjects design, does that matter?
The only within subject studies I have found using logistic regression are in medicine, the resource I quote above is from a lecture in psychology. My study is an educational evaluation which is more like a psychological study.
Can someone please enlighten me?
Answer by flemon99
Its hard to comment without knowing your specific predictors , but I don’t think anything you’ve said suggests that its inappropriate.
In regards to the independence of errors, note that we’re talking about the independence of errors of the DV, that which you’re predicting. In Logistic regression, this is usually a binary response (e.g. yes/no). This is what needs to be a between, not a within design (I assume participants cannot be both yes and no, so this should be okay).
Collinearity between the predictors might be an issue, which is your predictors being correlated with each other. If you are using a pre/post design, this is likely to be an issue (if you are using both as predictors, if not, it doesn’t matter so much. You could always easily use a single change score, post-pre score, to make it one value and get around this though).
It is commonly used in medicine, but not exclusively. Another common application I’ve come across is in relation to eye witness testimony/jury stuff (e.g. the likelihood of giving a guilty verdict). If your data is suitable though, and the analysis addresses your hypothesis, there isn’t any reason not to use it.
Hope that helps.
Give your answer to this question below!