README.md
In hobbitish1028/Logistic: Binomial logistic regression

Logistic

Perform binomial logistic regression

The goal of the package is to fit a binomial logistic regression via unpenalized maximum likelihood and to use the fitted model to make prediction on new data.

There are two functions in the package, Logreg and My_predict. The Logreg function fits binomial logistic models. The My_predict function provides the default fitting method for Logreg. It is similar to glm in the glm2 package, except for modifications to the computational method that provide more stable convergence.

We use adam-like algorithm (tuning-free optimization) to get the result. Besides, due to its theoretical superiority, it ususally performs better than SGD on data with high dimension (p is big).

We also give loss plot for user to check the convergence of optimization, while glm only give a warning or not to remind the user. Once we find the function diverge under some conditions, we can simply add to the parameter maxit to recover it.

Rcpp is used to speed up the loop part.

Although logistic regression is a strict convex question with a global maximum, since both models use stochastic optimization, they will converge to different solutions according to their stopping criterions and optimization rules. Thus we don't use benchmark here, and three digits of precision is already enough to make sure their convergence and consistency.

Use ?Logreg and ?My_predict to call help page.

And if you fail to call help page, please type .rs.restartR() and run it.