Logreg: Binomial Logistic regression

In hobbitish1028/Logistic: Binomial logistic regression

Description

Fit binomial logistic regression via unpenalized maximum likelihood. Use tuning-free optimization to get the result.

Usage

Logreg(X, y, maxit = 10000)

Arguments

X	input matrix, of dimension n (sample number) by p (variable number); each row is an observation vector.
y	response variable. Since we aim at binomial regression, it should be a vector with two levels (like a vector of 0 & 1, or a & b)
maxit	the Maximum number of iterations when we use optimization to estimate the parameeter; default is 10000. And the algorithm will stop according to its stopping criterion.

Value

A list containing the relevant data of regression result.

Examples

### install and library package (building vignetter may cost 1 minute)
#devtools::install_github("hobbitish1028/Logistic",build_vignettes = TRUE,force = TRUE)
#And if you fail to call help page after this, please type `.rs.restartR()` and run it.
library(Logistic)

### Generate training data
sigma<-4
set.seed(123)
n<-1e4
p<-1e2
mu1<-rnorm(p)
mu2<-rnorm(p)
X1<-matrix(mu1+rnorm(n*p,0,sigma),n,p,byrow = TRUE)
X2<-matrix(mu2+rnorm(n*p,0,sigma),n,p,byrow = TRUE)

### Train data
X<-rbind(X1,X2)
y<-rep(c(1,0),each=n)

### Fit model
fit<-Logreg(X,y)

### accuracy of train data
fit$accuracy

### check the convergence
plot(fit$loss)

### get the prediction of train data
result<-fit$prediction

### get the probability of each sample belonging to two category
result<-fit$P

### get the parameter of the logistic regression (including intercept)
result<-fit$x

browseVignettes("Logistic")

hobbitish1028/Logistic documentation built on Nov. 26, 2019, 8:58 a.m.