Given a matrix X of explanatory variables, a response y numeric variable, this function fits either a 'binomial' or 'multinomial' logistic regression model and returns a class object similar to sci-kit learn's object

Given a matrix X of explanatory variables, a response y numeric variable, this function fits either a 'binomial' or 'multinomial' logistic regression model and returns a class object similar to sci-kit learn's object

arid_logreg(X, y, regularization = NULL, lambda = NULL)

Arguments

X	(data frame): the explanatory variables matrix
y	(numeric): the response variable numeric vector
regularization	(character): what level of regularization to use in the model (default NULL)
lambda	(double): the regularization strength parameter to use (default: NULL)

Value

arid_logreg: a class object after fitting a 'binomial' or 'multinomial' logistic regression model

Examples

X <- matrix(rnorm(40 * 3), 40, 3)
y <- sample(c(0,1), 40, replace = TRUE)
arid_logreg(X, y)
#> $thisEnv
#> <environment: 0x000000004c316798>
#> 
#> $fit
#> function (X, y) 
#> {
#>     model <- NULL
#>     regularization <- regularization_
#>     lambda <- lambda_
#>     if (is.null(regularization)) {
#>         lambda <- 0
#>         model <- glmnet::glmnet(X, y, family = family, alpha = 1, 
#>             lambda = lambda)
#>     }
#>     else if (regularization == c("L1")) {
#>         model <- glmnet::glmnet(X, y, alpha = 1, family = family)
#>     }
#>     else if (regularization == c("L2")) {
#>         model <- glmnet::glmnet(X, y, alpha = 0, family = family)
#>     }
#>     else if (regularization == c("L1L2")) {
#>         model <- glmnet::glmnet(X, y, alpha = 0.5, family = family)
#>     }
#>     coefs <- get_coefs(X, y, model, lambda)
#>     set_coefs(coefs)
#>     return(model)
#> }
#> <bytecode: 0x000000004c308ac0>
#> <environment: 0x000000004c316798>
#> 
#> $predict
#> function (newx) 
#> {
#>     if (family == "binomial") {
#>         prob <- glmnet::predict.glmnet(model_, s = lambda_, newx = newx, 
#>             type = "response")
#>         pred <- ifelse(prob > 0.5, 1, 0)
#>         return(pred)
#>     }
#>     else {
#>         return("Only predictions for binomial logistic regression are available")
#>     }
#> }
#> <bytecode: 0x000000004c311360>
#> <environment: 0x000000004c316798>
#> 
#> $score
#> function () 
#> {
#>     model_$dev.ratio
#> }
#> <bytecode: 0x000000004c310870>
#> <environment: 0x000000004c316798>
#> 
#> $intercept_
#> [1] 0.004002721
#> 
#> $coef_
#> [1] -0.4014806 -0.5452708  0.0649694
#> 
#> attr(,"class")
#> [1] "arid_logreg"