predict.lb: Predict method for lb objects

In Libra: Linearized Bregman Algorithms for Generalized Linear Models

Predict method for lb objects

Description

Predict response variable for new data given a lb object

Usage

## S3 method for class 'lb'
predict(object, newx, t, type = c("fit", "coefficients"), ...)

Arguments

object	lb object
newx	New data matrix that each row is a data or a vector. If missing, type switched to coefficients
t	The parmeter for object to determin which coeffiecients used for prediction. Linear interpolation is used if t is not in object\$t. If missing, all the coeffiecients along the path is used to predict.
type	To predict response of newx or just fit coeffients on the path.
...	Additonal arguments for generic predict.

Details

The default plot uses the fraction of L1 norm as the x. For multinomial case, the sum of absolute values of different class's coefficients are caculated to represent each variable. The intercept term is not ploted

Value

A list containing t and other variables. For type="fit", the rediction response "fit" is returned. For "binomial", a vector of the probabilities for newx falling into class +1 is redurned. For "multinomial", a matrix with each column means the probabilities for newx falling into the corresponding class. If type="coefficients" coefficients "beta" and intercepts "a0" are returned.

Author(s)

Feng Ruan, Jiechao Xiong and Yuan Yao

Libra documentation built on April 11, 2022, 5:09 p.m.