R/predict.adapt.R

In wrbrooks/adapt: The adaptive lasso for linear regression

#' Make predictions from an adaptive lasso model.
#'
#' @param obj an adaptive lasso model
#' @param newx the covariate values to use for prediction
#'
#' @export
predict.adapt <-  function(obj, newx, lambda=obj$glmnet$lambda, type=c("link","response","coefficients","nonzero","class")) {
    pred.data = newx
    predictors = obj$predictors
    pred.data = pred.data[,predictors] %>% as.matrix %>% sweep(2, obj$meanx, '-') %>% sweep(2, obj$scale, '*')
    type = match.arg(type)

    if (obj$selectonly) {
        predictions = predict(obj$glm, newx)
    } else {
        predictions = predict(obj$glmnet, newx=pred.data, s=lambda, type=type)
    }

    return(predictions)
}