Minimum number of cases reqiured

[Chinonso Odebeatu]

Hi

I noticed in my data set that the number of cases is 34 (1.8%) while the number of non-cases is 1911 (98.2). Please, is it possible to run a logistic regression analysis on this sample?

[Leonardo Guizzetti]

Yes it is. Technically, you only need one case of each class of your dependent variable for logistic regression to estimate a model. Having said that, the low number of cases will limit the number of independent variables included in your regression model, and may impact both precision and bias of the estimates.

[Carole J. Wilson]

See Richard Williams’ notes on the issue:
https://www3.nd.edu/~rwilliam/stats3/RareEvents.pdf