Small sample size multivariant analysis

[Michelle Watson]

Hi guys,

I have a small sample size of around 350 of which around 20 or so had an event. I have around 12 variables that might be associated with the event and hoping to do multivariant analysis, some of which also have small number around 10. I m keen to know 1 particular variable and if it has significant impact on the event taking consideration into the others.

Is exact logistic regression my best option here?

Thank you!

[Leonardo Guizzetti]

Hi Michelle, welcome to the forum. Please take some time to review the FAQ and in particular, how to effectively ask questions here. Your post is confusing to me for several reasons.

1) You speak of events and then raise logistic regression. Are you certain that time-to-event analysis is not more appropriate?

2) "Multivariant" is not a statistical term, but multivariable and multivariate are. Multivariate is used in the context when there are more than out outcome (more than 1 variable on the left side of the regression equation, for example). Multivariable is when there is more than one predictor variable (more than 1 variable on the right side of the regression equation, for example).

I will assume that you intended to ask about multivariable logistic regression. If that's the case, then you have a relatively small number of events (20/350). 20 is the lesser number of events and non-events, and this number governs the effective sample size and therefore limits how many predictors you can include in the model. Various rules of thumb exist, but 10 events-to-1 parameter estimate is a reasonable place to start, so you can effectively include a max of 2 variables in your model. My advice would be the example univariable logistic regression models, and examine each variable one at a time, and report on those.