plot_ace: Plot predictions of the Additive Causal Expansion model
In mazphilip/GPspline: Additive Causal Expansion Model

Description Usage Arguments Details Value

View source: R/plot_ace.R

This function plots the response surface or the marginal effect fitted by the ace package around the training sample. For non-binary treatments, the method can display the interaction of the treatment with one confounder. The package plots:

for binary treatments, a 2D plot with X[, Xcol] on the x-axis. This plots,
- if marginal = FALSE: the treatment and control reponse surfaces.
- if marginal = TRUE: the treatment effect.
For non-binary (continuous and -for now- categorial) treatments the method can display a 2D contour or a 3D surface plot with the interaction between the treatment and a specified confounder through Xcol. If no confounder is selected, a simple 2D plot with the treatment on the x-axis is returned. Depending on marginal the levels correspond either to the response surface (if FALSE, default) or the marginal treatment effect/slope of response surface (if TRUE). The different confounder plots can be accessed using:
- if plot3D = TRUE and Xcol specified: A 3D surface using plotly (if plotly installed, otherwise plot3D defaults to FALSE). The treatment variable and the selected X[, Xcol] are shown on the x- and y-axis, the outcome surface on the z-axis.
- if plot3D = FALSE and Xcol specified: the treatment effect for different treatment levels using a (2D) contour plot with the treatment variable and the selected X[, Xcol] on the x- and y-axis.
- if no Xcol specified: A 2D plot with treatment levels on the x- and the outcome on the y-axis.

1
2
3

plot_ace(object, Xcol, marginal = FALSE, plot3D = FALSE,
  show.observations = TRUE, Xlim = c(-1, 1), Xstep = 0.1, Zlim = c(-1,
  1), Zstep = 0.1, truefun, ...)

`object`	An `ace` object as generated by the ace.train function.
`Xcol`	A numeric. If missing plots the 2D graph of Y on Z.
`marginal`	A logical statement, default: `FALSE`. If `TRUE`, plots the marginal surface of the response surface Y with respect to the basis expanded model Z using average X.
`plot3D`	A logical statement whether a three-dimensional plot should be produced. This is only valid if an X-column is selected and the plotly package is installed.
`show.observations`	A logical scalar that determines
`Xlim`	A two-dimensional vector with thelimits of the confounder dimension.
`Xstep`	A scalar for the step-size of the confounder dimension.
`Zlim`	A two-dimensional vector with thelimits of the treatment dimension.
`Zstep`	A scalar for the step-size of the treatment dimension.
`truefun`	A function handle for simulations.
`...`	Arguments handed to plotting function (`plot`, `contour`, `plotly`).