Combine varpred objects

Combines and plots comparison plots for more than two named varpred objects.

Usage

combinevarpred(
  vlist,
  lnames = NULL,
  plotit = FALSE,
  addmarginals = FALSE,
  margindex,
  ...
)

Arguments

vlist

a list of varpred objects.

lnames

a character vector specifying the name(s) of the vlist objects. Useful when faceted comparison is needed. See examples.

plotit

logical. If TRUE a plot is returned, otherwise a varpred object.

addmarginals

logical. If TRUE the mean of the estimates is added to the plot.

margindex

an integer vector indexing the vlist. Useful if particular vlist needs to be averaged together in the addmarginals.

...

additional arguments passed to plot.varpred.

Value

a varpred object or a plot.

See also

varpred

Examples

# Set theme for ggplot. Comment out if not needed
library(ggplot2)
varpredtheme()
set.seed(911)
# Simulate binary outcome data with two predictors
steps <- 500
N <- 100
b0 <- 2
b_age <- -1.5
b_income <- 1.8
min_age <- 18
age <- min_age + rnorm(N, 0, 1)
min_income <- 15
income <- min_income + rnorm(N, 0, 1)
eta <- b0 + age*b_age + income*b_income
status <- rbinom(N, 1, plogis(eta))
df <- data.frame(status, age, income)

# Fit model
mod <- glm(status ~ age + income, df, family=binomial())

# Effect plots
## Mean-based
ef_mean <- varpred(mod, "age", steps=steps, bias.adjust="none", modelname="mean-based")
## Observed-value-based
ef_observed <- varpred(mod, "age", steps=steps, bias.adjust="observed", modelname="observed-value")
## Combine all the effect estimates
ef <- combinevarpred(list(ef_mean, ef_observed))
print(plot(ef)
  + scale_color_brewer(palette = "Dark2")
)
#> Scale for 'colour' is already present. Adding another scale for 'colour',
#> which will replace the existing scale.


# Prediction plots
## Mean-based
pred_mean <- varpred(mod, "age", isolate=FALSE
  , steps=steps, bias.adjust="none"
  , modelname="mean-based"
)
## Observed-value-based
pred_observed <- varpred(mod, "age", isolate=FALSE
  , steps=steps, bias.adjust="observed"
  , modelname="observed-value"
)
## Combine all the prediction estimates
### With plotit=TRUE no need to plot
pred <- combinevarpred(list(pred_mean, pred_observed), plotit=TRUE)
print(pred
  + scale_color_brewer(palette = "Dark2")
)
#> Scale for 'colour' is already present. Adding another scale for 'colour',
#> which will replace the existing scale.