1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/compute_model_prediction.R
\name{compute_model_prediction}
\alias{compute_model_prediction}
\alias{compute_smooth}
\title{Create a model of a data set and compute predictions.}
\usage{
compute_model_prediction(
x,
formula,
...,
model = NULL,
se = FALSE,
level = 0.95,
n = 80L,
domain = NULL,
method
)
compute_smooth(x, formula, ..., span = 0.75, se = FALSE)
}
\arguments{
\item{x}{Dataset-like object to model and predict. Built-in methods for data
frames, grouped data frames and ggvis visualisations.}
\item{formula}{Formula passed to modelling function. Can use any variables
from data.}
\item{...}{arguments passed on to \code{model} function}
\item{model}{Model fitting function to use - it must support R's standard
modelling interface, taking a formula and data frame as input, and
returning predictions with \code{\link{predict}}. If not supplied, will use
\code{\link{loess}} for <= 1000 points, otherwise it will use
\code{\link[mgcv]{gam}}. Other modelling functions that will work include
\code{\link{lm}}, \code{\link{glm}} and \code{\link[MASS]{rlm}}.}
\item{se}{include standard errors in output? Requires appropriate method of
\code{predict_grid}, since the interface for returning predictions with
standard errors is not consistent acrossing modelling frameworks.}
\item{level}{the confidence level of the standard errors.}
\item{n}{the number of grid points to use in the prediction}
\item{domain}{If \code{NULL} (the default), the domain of the predicted
values will be the same as the domain of the prediction variable in the
data. It can also be a two-element numeric vector specifying the min and
max.}
\item{method}{Deprecated. Please use \code{model} instead.}
\item{span}{Smoothing span used for loess model.}
}
\value{
A data frame with columns: \item{\code{resp_}}{regularly spaced grid
of \code{n} locations} \item{\code{pred_}}{predicted value from model}
\item{\code{pred_lwr_} and \code{pred_upr_}}{upper and lower bounds of
confidence interval (if \code{se = TRUE})} \item{\code{pred_se_}}{the
standard error (width of the confidence interval) (if \code{se = TRUE})}
}
\description{
Fit a 1d model, then compute predictions and (optionally) standard errors
over an evenly spaced grid.
}
\details{
\code{compute_model_prediction} fits a model to the data and makes
predictions with it. \code{compute_smooth} is a special case of model
predictions where the model is a smooth loess curve whose smoothness is
controlled by the \code{span} parameter.
}
\examples{
# Use a small value of n for these examples
mtcars \%>\% compute_model_prediction(mpg ~ wt, n = 10)
mtcars \%>\% compute_model_prediction(mpg ~ wt, n = 10, se = TRUE)
mtcars \%>\% group_by(cyl) \%>\% compute_model_prediction(mpg ~ wt, n = 10)
# compute_smooth defaults to loess
mtcars \%>\% compute_smooth(mpg ~ wt)
# Override model to suppress message or change approach
mtcars \%>\% compute_model_prediction(mpg ~ wt, n = 10, model = "loess")
mtcars \%>\% compute_model_prediction(mpg ~ wt, n = 10, model = "lm")
# Set the domain manually
mtcars \%>\%
compute_model_prediction(mpg ~ wt, n = 20, model = "lm", domain = c(0, 8))
# Plot the results
mtcars \%>\% compute_model_prediction(mpg ~ wt) \%>\%
ggvis(~pred_, ~resp_) \%>\%
layer_paths()
mtcars \%>\% ggvis() \%>\%
compute_model_prediction(mpg ~ wt) \%>\%
layer_paths(~pred_, ~resp_)
}
|
