binaryloess: Using loess to Check Functional Form for Logistic Regression

In kyotil: Utility Functions for Statistical Analysis Report Generation and Monte Carlo Studies

Using loess to Check Functional Form for Logistic Regression

Description

This function plots a smoothed line of how the average value of Y changes with X in order to check functional form for logistic regression.

Usage

binaryloess(x, y, scale = c("logit", "linear"), span = 0.7, weights = NULL, ...)

Arguments

x	tbdx
y	tbdy
scale	tbdscale
span	smoothing parameter, passed to loess. If less than 1, the neighbourhood includes proportion a of the points. If greater than 1, all points are used, with the maximum distance assumed to be a^(1/p) times the actual maximum distance for p explanatory variables. Missing records are removed first.
weights	sampling weights, passed to loess
...	passed to plotting function

Details

This function comes from Jonathan Bartlett ()https://thestatsgeek.com/2014/09/13/checking-functional-form-in-logistic-regression-using-loess/).

Examples

set.seed(1234)
n <- 1000
x <- rnorm(n)
xb <- -2+x
pr <- exp(xb)/(1+exp(xb))
y=rbern(n, pr)

par(mfrow=c(1,2))
binaryloess(x, y, scale = "logit", span = 0.7, weights = NULL, ylab="logit(p)")
binaryloess(x, y, scale = "linear", span = 0.7, weights = NULL, ylab="prob")

kyotil documentation built on Nov. 3, 2024, 9:08 a.m.