epdiff1: Explain difference between predictions from a model

In CollinErickson/ExplainPrediction2: Explain Difference in Predictions From a Model

Description

Explain difference between predictions from a model

Usage

epdiff1(
  mod,
  x1,
  x2,
  d,
  predictfunc,
  predictfuncwithmod,
  name1,
  name2,
  namesfromcol
)

Arguments

mod	Model
x1	Input 1. Single row data frame or matrix.
x2	Input 2. Single row data frame or matrix.
d	Max number of input dimensions to use.
predictfunc	Func to use to predict. Defaults to 'predict'. Takes in model and data.
predictfuncwithmod	Func to use to predict. Only takes in data.
name1	Name x1
name2	Name for x2
namesfromcol	Column of x1/x2 to get name from

Value

ggplot object showing effect

Examples

# gbm
m1 <- gbm::gbm(Petal.Width ~ Sepal.Length + Sepal.Width, data=iris, distribution='gaussian')
x1 <- iris[1,]
x2 <- iris[2,]
epdiff1(m1, x1, x2)
epdiff1(m1, iris[12,], iris[20,])
epdiff1(m1, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),])
epdiff1(m1, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),], predictfunc=predict)
epdiff1(m1, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),],
    predictfuncwithmod=function(xx){suppressMessages(predict(m1, xx))})
epdiff1(NA, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),],
    predictfuncwithmod=function(xx){suppressMessages(predict(m1, xx))})
epdiff1(m1, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),], name1='aa', name2='bb')
epdiff1(m1, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),], namesfromcol="Species")

# Linear model
mod_lm <- lm(Petal.Width ~ Sepal.Length + Sepal.Width, data=iris)
epdiff1(mod_lm, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),])
epdiff1(mod_lm, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),], d=3)
mod_lm2 <- lm(Petal.Width ~ Petal.Length + Sepal.Width, data=iris)
epdiff1(mod_lm2, iris[sample(1:nrow(iris), 1),], iris[sample(1:nrow(iris), 1),])

CollinErickson/ExplainPrediction2 documentation built on Dec. 17, 2021, 3:02 p.m.