Logistic Regression

[Rahim Abdool]

I am fairly new to Statistics. I have a large datatset where N=50,000. I have a dichotomous outcome variable Death (0 = Alive 1 = Die) and a categorical variable Weight (0=Normal 1= Obese 2 = Underweight). I am investigating whether weight contribute to an early death by running a logistic regression. I am wondering whether there is a way to check the lineartiy assumption? The Box-Tidwell test does not provide a p-value in this instance. Any advice will be much appreciated. Many Thanks.

[Felix Bittmann]

I am not sure what you mean by linearity assumption. The outcome is binary and the dependent variable has 3 distinct, non-metric categories. I think you do not need to check for any more assumptions.

[Carlo Lazzaro]

Rahim:
welcome to this forum.
I do share Felix's concern about what you mean by linearity.
My guess-work is that you're interested in turning point, but this is unfeasible with a categorical predictor.
In addition, and much more substantively, as an acid test I've just performed a basic PubMed search (https://pubmed.ncbi.nlm.nih.gov/?term=weight+death&sort=date) that returned 62,980 results.
Far from me any intent to be disrespectful, blunt or outspoken, but I'm afraid you will go nowhere with a simple logistic regression applied to such a complex issue.