ceplot: Interactive conditional expectation plot

In markajoc/condvis: Conditional Visualization for Statistical Models

Description

Creates an interactive conditional expectation plot, which consists of two main parts. One part is a single plot depicting a section through a fitted model surface, or conditional expectation. The other part shows small data summaries which give the current condition, which can be altered by clicking with the mouse.

Usage

ceplot(data, model, response = NULL, sectionvars = NULL,
  conditionvars = NULL, threshold = NULL, lambda = NULL,
  distance = c("euclidean", "maxnorm"), type = c("default", "separate",
  "shiny"), view3d = FALSE, Corder = "default", selectortype = "minimal",
  conf = FALSE, probs = FALSE, col = "black", pch = NULL,
  residuals = FALSE, xsplotpar = NULL, modelpar = NULL,
  xcplotpar = NULL)

Arguments

data	A dataframe containing the data to plot
model	A model object, or list of model objects
response	Character name of response in data
sectionvars	Character name of variable(s) from data on which to take a section, can be of length 1 or 2.
conditionvars	Character names of conditioning variables from data. These are the predictors which we can set to single values in order to produce a section. Can be a list of vectors of length 1 or 2. Can be a character vector, which is then paired up using arrangeC. If NULL, an attempt will be made to extract all variable names which are not response or sectionvars from model, and these will be arranged using arrangeC.
threshold	This is a threshold distance. Points further than threshold away from the current section will not be visible. Passed to similarityweight.
lambda	A constant to multiply by number of factor mismatches in constructing a general dissimilarity measure. If left NULL, behaves as though lambda is set greater than threshold, and so only observations whose factor levels match the current section are visible. Passed to similarityweight.
distance	A character vector describing the type of distance measure to use, either "euclidean" (default) or "maxnorm".
type	This specifies the type of interactive plot. "default" places everything on one device. "separate" places condition selectors on one device and the section on another. (These two options require XQuartz on OS X). "shiny" produces a Shiny application.
view3d	Logical; if TRUE plots a three-dimensional regression surface if possible.
Corder	Character name for method of ordering conditioning variables. See arrangeC.
selectortype	Type of condition selector plots to use. Must be "minimal" if type is "default". If type is "separate", can be "pcp" (see plotxc.pcp) or "full" (see plotxc.full).
conf	Logical; if TRUE plots confidence bounds (or equivalent) for models which provide this.
probs	Logical; if TRUE, shows predicted class probabilities instead of just predicted classes. Only available if S specifies two numeric predictors and the model's predict method provides this.
col	Colour for observed data.
pch	Plot symbols for observed data.
residuals	Logical; if TRUE, plots a residual versus predictor plot instead of the usual scale of raw response.
xsplotpar	Plotting parameters for section visualisation as a list, passed to plotxs. Can specify xlim, ylim.
modelpar	Plotting parameters for models as a list, passed to plotxs. Not used.
xcplotpar	Plotting parameters for condition selector plots as a list, passed to plotxc. Can specify col for highlighting current section, cex, and trim (see plotxc).

References

O'Connell M, Hurley CB and Domijan K (2017). “Conditional Visualization for Statistical Models: An Introduction to the condvis Package in R.”Journal of Statistical Software, 81(5), pp. 1-20. <URL:http://dx.doi.org/10.18637/jss.v081.i05>.

See Also

condtour, similarityweight

Examples

## Not run: 
## Example 1: Multivariate regression, xs one continuous predictor

mtcars$cyl <- as.factor(mtcars$cyl)

library(mgcv)
model1 <- list(
    quadratic = lm(mpg ~ cyl + hp + wt + I(wt^2), data = mtcars),
    additive = mgcv::gam(mpg ~ cyl + hp + s(wt), data = mtcars))

conditionvars1 <- list(c("cyl", "hp"))

ceplot(data = mtcars, model = model1, response = "mpg", sectionvars = "wt",
  conditionvars = conditionvars1, threshold = 0.3, conf = T)

## Example 2: Binary classification, xs one categorical predictor

mtcars$cyl <- as.factor(mtcars$cyl)
mtcars$am <- as.factor(mtcars$am)

library(e1071)
model2 <- list(
  svm = svm(am ~ mpg + wt + cyl, data = mtcars, family = "binomial"),
  glm = glm(am ~ mpg + wt + cyl, data = mtcars, family = "binomial"))

ceplot(data = mtcars, model = model2, sectionvars = "wt", threshold = 1,
  type = "shiny")

## Example 3: Multivariate regression, xs both continuous

mtcars$cyl <- as.factor(mtcars$cyl)
mtcars$gear <- as.factor(mtcars$gear)

library(e1071)
model3 <- list(svm(mpg ~ wt + qsec + cyl + hp + gear,
  data = mtcars, family = "binomial"))

conditionvars3 <- list(c("cyl","gear"), "hp")

ceplot(data = mtcars, model = model3, sectionvars = c("wt", "qsec"),
  threshold = 1, conditionvars = conditionvars3)

ceplot(data = mtcars, model = model3, sectionvars = c("wt", "qsec"),
    threshold = 1, type = "separate", view3d = T)

## Example 4: Multi-class classification, xs both categorical

mtcars$cyl <- as.factor(mtcars$cyl)
mtcars$vs <- as.factor(mtcars$vs)
mtcars$am <- as.factor(mtcars$am)
mtcars$gear <- as.factor(mtcars$gear)
mtcars$carb <- as.factor(mtcars$carb)

library(e1071)
model4 <- list(svm(carb ~ ., data = mtcars, family = "binomial"))

ceplot(data = mtcars, model = model4, sectionvars = c("cyl", "gear"),
  threshold = 3)

## Example 5: Multi-class classification, xs both continuous

data(wine)
wine$Class <- as.factor(wine$Class)
library(e1071)

model5 <- list(svm(Class ~ ., data = wine, probability = TRUE))

ceplot(data = wine, model = model5, sectionvars = c("Hue", "Flavanoids"),
  threshold = 3, probs = TRUE)

ceplot(data = wine, model = model5, sectionvars = c("Hue", "Flavanoids"),
  threshold = 3, type = "separate")

ceplot(data = wine, model = model5, sectionvars = c("Hue", "Flavanoids"),
  threshold = 3, type = "separate", selectortype = "pcp")

## Example 6: Multi-class classification, xs with one categorical predictor,
##            and one continuous predictor.

mtcars$cyl <- as.factor(mtcars$cyl)
mtcars$carb <- as.factor(mtcars$carb)

library(e1071)
model6 <- list(svm(cyl ~ carb + wt + hp, data = mtcars, family = "binomial"))

ceplot(data = mtcars, model = model6, threshold = 1, sectionvars = c("carb",
  "wt"), conditionvars = "hp")

## End(Not run)

markajoc/condvis documentation built on May 21, 2019, 11:48 a.m.