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
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
|
% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/interactions.R
\name{step_interact}
\alias{step_interact}
\title{Create Interaction Variables}
\usage{
step_interact(
recipe,
terms,
role = "predictor",
trained = FALSE,
objects = NULL,
sep = "_x_",
skip = FALSE,
id = rand_id("interact")
)
}
\arguments{
\item{recipe}{A recipe object. The step will be added to the
sequence of operations for this recipe.}
\item{terms}{A traditional R formula that contains interaction
terms. This can include \code{.} and selectors. See \code{\link[=selections]{selections()}}
for more details, and consider using \code{\link[tidyselect:starts_with]{tidyselect::starts_with()}} when
dummy variables have been created.}
\item{role}{For model terms created by this step, what analysis role should
they be assigned? By default, the new columns created by this step from
the original variables will be used as \emph{predictors} in a model.}
\item{trained}{A logical to indicate if the quantities for
preprocessing have been estimated.}
\item{objects}{A list of \code{terms} objects for each
individual interaction.}
\item{sep}{A character value used to delineate variables in an
interaction (e.g. \code{var1_x_var2} instead of the more
traditional \code{var1:var2}).}
\item{skip}{A logical. Should the step be skipped when the
recipe is baked by \code{\link[=bake]{bake()}}? While all operations are baked
when \code{\link[=prep]{prep()}} is run, some operations may not be able to be
conducted on new data (e.g. processing the outcome variable(s)).
Care should be taken when using \code{skip = TRUE} as it may affect
the computations for subsequent operations.}
\item{id}{A character string that is unique to this step to identify it.}
}
\value{
An updated version of \code{recipe} with the new step added to the
sequence of any existing operations.
}
\description{
\code{step_interact} creates a \emph{specification} of a recipe
step that will create new columns that are interaction terms
between two or more variables.
}
\details{
\code{step_interact} can create interactions between
variables. It is primarily intended for \strong{numeric data};
categorical variables should probably be converted to dummy
variables using \code{\link[=step_dummy]{step_dummy()}} prior to being used for
interactions.
Unlike other step functions, the \code{terms} argument should
be a traditional R model formula but should contain no inline
functions (e.g. \code{log}). For example, for predictors
\code{A}, \code{B}, and \code{C}, a formula such as
\code{~A:B:C} can be used to make a three way interaction
between the variables. If the formula contains terms other than
interactions (e.g. \code{(A+B+C)^3}) only the interaction terms
are retained for the design matrix.
The separator between the variables defaults to "\verb{_x_}" so
that the three way interaction shown previously would generate a
column named \code{A_x_B_x_C}. This can be changed using the
\code{sep} argument.
When dummy variables are created and are used in interactions,
selectors can help specify the interactions succinctly. For
example, suppose a factor column \code{X} gets converted to dummy
variables \code{x_2}, \code{x_3}, ..., \code{x_6} using \code{\link[=step_dummy]{step_dummy()}}. If
you wanted an interaction with numeric column \code{z}, you could
create a set of specific interaction effects (e.g.
\code{x_2:z + x_3:z} and so on) or you could use
\code{starts_with("x_"):z}. When \code{\link[=prep]{prep()}} evaluates this step,
\code{starts_with("x_")} resolves to \code{(x_2 + x_3 + x_4 + x_5 + x_6)}
so that the formula is now \code{(x_2 + x_3 + x_4 + x_5 + x_6):z} and
all two-way interactions are created.
}
\section{Tidying}{
When you \code{\link[=tidy.recipe]{tidy()}} this step, a tibble with column
\code{terms} (the interaction effects) is returned.
}
\section{Case weights}{
The underlying operation does not allow for case weights.
}
\examples{
\dontshow{if (rlang::is_installed("modeldata")) (if (getRversion() >= "3.4") withAutoprint else force)(\{ # examplesIf}
data(penguins, package = "modeldata")
penguins <- penguins \%>\% na.omit()
rec <- recipe(flipper_length_mm ~ ., data = penguins)
int_mod_1 <- rec \%>\%
step_interact(terms = ~ bill_depth_mm:bill_length_mm)
# specify all dummy variables succinctly with `starts_with()`
int_mod_2 <- rec \%>\%
step_dummy(sex, species, island) \%>\%
step_interact(terms = ~ body_mass_g:starts_with("species"))
int_mod_1 <- prep(int_mod_1, training = penguins)
int_mod_2 <- prep(int_mod_2, training = penguins)
dat_1 <- bake(int_mod_1, penguins)
dat_2 <- bake(int_mod_2, penguins)
names(dat_1)
names(dat_2)
tidy(int_mod_1, number = 1)
tidy(int_mod_2, number = 2)
\dontshow{\}) # examplesIf}
}
|
