I hope that you understand what "exact" means in exact logistic regression, and are sure that you want to employ that technique. It certainly does not have the naive interpretation that it gives you exact answers, whereas the standard logistic regression only gives you approximate/inferior answers. A better approach with sparse data is to use some type of shrinkage estimation so that the model parameters are not highly unstable. See Sander Greenland's papers (two part paper) on Bayesian perspectives for epidemiological research (Intl J Epi, 2006).

Also, FWIW, the elrm package is, in fact, available on CRAN-- at least as of this moment.

Neither exact nor shrinkage estimators (like Firth's logistic regression) are likely to help you here. I would recommend you not attempt to adjust for more than 2 or 3 regressors in a multivariate model, but resign yourself to a series of bivariate assessments of association.

I have experienced in running R for exact logistic regresssion using elrm package. If we have more than one regressor, elrm can not run regressors simultaneously, but it only runs one regressor. Suppose we have two regressors, indep1 and indep2 and one dependent variable (depvar) --> test

Hi dear colleagues!

What is R package to continue elrm package for exact logistic regression? It is seem erlm is not supported by CRAN.

Hi Amir,

I haven't been able to find an alternative package for exact logistic regression that is still supported, but you can still download elrm from the archive here:

https://cran.r-project.org/src/contrib/Archive/elrm

Dear Madison

Thanks for your attention. installing elrm via Rtools was successful from mentioned archive.

Best Wishes.

You can install elrm via: devtools::install_version("elrm", "1.2.2")