gmodel: Graph the function implicit in a model

In dtkaplan/statisticalModeling: Functions for Teaching Statistical Modeling

Description

Currently, this is the same as fmodel. I think the name gmodel makes more sense, and I want to be able to add additional functionality (such as including data points on the plot) without breaking fmodel.

Usage

gmodel(model = NULL, formula = NULL, data = NULL, nlevels = 3,
  at = list(), prob_of = NULL, intervals = c("none", "confidence",
  "prediction"), post_transform = NULL, ...)

Arguments

model	the model to display graphically
formula	setting the y ~ x + color variables
data	optional data set from which to extract levels for explanatory variables
nlevels	how many levels to display for those variables shown at discrete levels
at	named list giving specific values at which to hold the variables. You can accomplish this without forming a list by using .... See examples.
prob_of	if to show probability of a given level of the output, name the class here as a character string.
intervals	show confidence or prediction intervals: values "none", "confidence", "prediction"
post_transform	a scalar transformation and new name for the response variable, e.g. post_transform = c(price = exp) to undo a log transformation of price.
...	specific values for explantory variables and/or arguments to predict()

Details

#' Plot out model values

Often you will want to show some data along with the model functions. You can do this with 'ggplot2::geom_point()' making sure to set the data argument to be a data frame with the cases you want to plot.

Examples

## Not run: 
mod1 <- lm(wage ~ age * sex + sector, data = mosaicData::CPS85)
fmodel(mod1)
fmodel(mod1, ~ sector + sex + age) # not necessarily a good ordering
# show the data used for fitting along with the model
fmodel(mod1, ~ age + sex + sector, nlevels = 8) + 
  ggplot2::geom_point(data = mosaicData::CPS85, alpha = 0.1)
require(ggplot2)
fmodel(mod1, ~ age + sex + sector, nlevels = 8) + 
  geom_point(data = mosaicData::CPS85, alpha = 0.1) +
  ylim(0, 20)
mod2 <- lm(log(wage) ~ age + sex + sector, data = mosaicData::CPS85)
fmodel(mod2, post_transform = c(wage = exp)) # undo the log in the display
mod3 <- glm(married == "Married" ~ age + sex * sector,
            data = mosaicData::CPS85, family = "binomial")
fmodel(mod3, type = "response")
# Adding the raw data requires an as.numeric() trick when it's TRUE/FALSE
fmodel(mod3, ~ age + sex + sector, nlevels = 10, type = "response") + 
  geom_point(data = mosaicData::CPS85, 
  aes(x = age, y = as.numeric(married == "Married")), alpha = .1)

## End(Not run)

dtkaplan/statisticalModeling documentation built on May 15, 2019, 5:23 p.m.