geom_smooth.Rd

% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/geom-smooth.R, R/stat-smooth.R
\name{geom_smooth}
\alias{geom_smooth}
\alias{stat_smooth}
\title{Smoothed conditional means}
\usage{
geom_smooth(
  mapping = NULL,
  data = NULL,
  stat = "smooth",
  position = "identity",
  ...,
  method = NULL,
  formula = NULL,
  se = TRUE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_smooth(
  mapping = NULL,
  data = NULL,
  geom = "smooth",
  position = "identity",
  ...,
  method = NULL,
  formula = NULL,
  se = TRUE,
  n = 80,
  span = 0.75,
  fullrange = FALSE,
  xseq = NULL,
  level = 0.95,
  method.args = list(),
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)
}
\arguments{
\item{mapping}{Set of aesthetic mappings created by \code{\link[=aes]{aes()}}. If specified and
\code{inherit.aes = TRUE} (the default), it is combined with the default mapping
at the top level of the plot. You must supply \code{mapping} if there is no plot
mapping.}

\item{data}{The data to be displayed in this layer. There are three
options:

If \code{NULL}, the default, the data is inherited from the plot
data as specified in the call to \code{\link[=ggplot]{ggplot()}}.

A \code{data.frame}, or other object, will override the plot
data. All objects will be fortified to produce a data frame. See
\code{\link[=fortify]{fortify()}} for which variables will be created.

A \code{function} will be called with a single argument,
the plot data. The return value must be a \code{data.frame}, and
will be used as the layer data. A \code{function} can be created
from a \code{formula} (e.g. \code{~ head(.x, 10)}).}

\item{position}{A position adjustment to use on the data for this layer. This
can be used in various ways, including to prevent overplotting and
improving the display. The \code{position} argument accepts the following:
\itemize{
\item The result of calling a position function, such as \code{position_jitter()}.
This method allows for passing extra arguments to the position.
\item A string naming the position adjustment. To give the position as a
string, strip the function name of the \code{position_} prefix. For example,
to use \code{position_jitter()}, give the position as \code{"jitter"}.
\item For more information and other ways to specify the position, see the
\link[=layer_positions]{layer position} documentation.
}}

\item{...}{Other arguments passed on to \code{\link[=layer]{layer()}}'s \code{params} argument. These
arguments broadly fall into one of 4 categories below. Notably, further
arguments to the \code{position} argument, or aesthetics that are required
can \emph{not} be passed through \code{...}. Unknown arguments that are not part
of the 4 categories below are ignored.
\itemize{
\item Static aesthetics that are not mapped to a scale, but are at a fixed
value and apply to the layer as a whole. For example, \code{colour = "red"}
or \code{linewidth = 3}. The geom's documentation has an \strong{Aesthetics}
section that lists the available options. The 'required' aesthetics
cannot be passed on to the \code{params}. Please note that while passing
unmapped aesthetics as vectors is technically possible, the order and
required length is not guaranteed to be parallel to the input data.
\item When constructing a layer using
a \verb{stat_*()} function, the \code{...} argument can be used to pass on
parameters to the \code{geom} part of the layer. An example of this is
\code{stat_density(geom = "area", outline.type = "both")}. The geom's
documentation lists which parameters it can accept.
\item Inversely, when constructing a layer using a
\verb{geom_*()} function, the \code{...} argument can be used to pass on parameters
to the \code{stat} part of the layer. An example of this is
\code{geom_area(stat = "density", adjust = 0.5)}. The stat's documentation
lists which parameters it can accept.
\item The \code{key_glyph} argument of \code{\link[=layer]{layer()}} may also be passed on through
\code{...}. This can be one of the functions described as
\link[=draw_key]{key glyphs}, to change the display of the layer in the legend.
}}

\item{method}{Smoothing method (function) to use, accepts either
\code{NULL} or a character vector, e.g. \code{"lm"}, \code{"glm"}, \code{"gam"}, \code{"loess"}
or a function, e.g. \code{MASS::rlm} or \code{mgcv::gam}, \code{stats::lm}, or \code{stats::loess}.
\code{"auto"} is also accepted for backwards compatibility.  It is equivalent to
\code{NULL}.

For \code{method = NULL} the smoothing method is chosen based on the
size of the largest group (across all panels). \code{\link[stats:loess]{stats::loess()}} is
used for less than 1,000 observations; otherwise \code{\link[mgcv:gam]{mgcv::gam()}} is
used with \code{formula = y ~ s(x, bs = "cs")} with \code{method = "REML"}. Somewhat anecdotally,
\code{loess} gives a better appearance, but is \eqn{O(N^{2})}{O(N^2)} in memory,
so does not work for larger datasets.

If you have fewer than 1,000 observations but want to use the same \code{gam()}
model that \code{method = NULL} would use, then set
\verb{method = "gam", formula = y ~ s(x, bs = "cs")}.}

\item{formula}{Formula to use in smoothing function, eg. \code{y ~ x},
\code{y ~ poly(x, 2)}, \code{y ~ log(x)}. \code{NULL} by default, in which case
\code{method = NULL} implies \code{formula = y ~ x} when there are fewer than 1,000
observations and \code{formula = y ~ s(x, bs = "cs")} otherwise.}

\item{se}{Display confidence interval around smooth? (\code{TRUE} by default, see
\code{level} to control.)}

\item{na.rm}{If \code{FALSE}, the default, missing values are removed with
a warning. If \code{TRUE}, missing values are silently removed.}

\item{orientation}{The orientation of the layer. The default (\code{NA})
automatically determines the orientation from the aesthetic mapping. In the
rare event that this fails it can be given explicitly by setting \code{orientation}
to either \code{"x"} or \code{"y"}. See the \emph{Orientation} section for more detail.}

\item{show.legend}{logical. Should this layer be included in the legends?
\code{NA}, the default, includes if any aesthetics are mapped.
\code{FALSE} never includes, and \code{TRUE} always includes.
It can also be a named logical vector to finely select the aesthetics to
display.}

\item{inherit.aes}{If \code{FALSE}, overrides the default aesthetics,
rather than combining with them. This is most useful for helper functions
that define both data and aesthetics and shouldn't inherit behaviour from
the default plot specification, e.g. \code{\link[=borders]{borders()}}.}

\item{geom, stat}{Use to override the default connection between
\code{geom_smooth()} and \code{stat_smooth()}. For more information about overriding
these connections, see how the \link[=layer_stats]{stat} and \link[=layer_geoms]{geom}
arguments work.}

\item{n}{Number of points at which to evaluate smoother.}

\item{span}{Controls the amount of smoothing for the default loess smoother.
Smaller numbers produce wigglier lines, larger numbers produce smoother
lines. Only used with loess, i.e. when \code{method = "loess"},
or when \code{method = NULL} (the default) and there are fewer than 1,000
observations.}

\item{fullrange}{If \code{TRUE}, the smoothing line gets expanded to the range of the plot,
potentially beyond the data. This does not extend the line into any additional padding
created by \code{expansion}.}

\item{xseq}{A numeric vector of values at which the smoother is evaluated.
When \code{NULL} (default), \code{xseq} is internally evaluated as a sequence of \code{n}
equally spaced points for continuous data.}

\item{level}{Level of confidence interval to use (0.95 by default).}

\item{method.args}{List of additional arguments passed on to the modelling
function defined by \code{method}.}
}
\description{
Aids the eye in seeing patterns in the presence of overplotting.
\code{geom_smooth()} and \code{stat_smooth()} are effectively aliases: they
both use the same arguments. Use \code{stat_smooth()} if you want to
display the results with a non-standard geom.
}
\details{
Calculation is performed by the (currently undocumented)
\code{predictdf()} generic and its methods.  For most methods the standard
error bounds are computed using the \code{\link[=predict]{predict()}} method -- the
exceptions are \code{loess()}, which uses a t-based approximation, and
\code{glm()}, where the normal confidence interval is constructed on the link
scale and then back-transformed to the response scale.
}
\section{Orientation}{

This geom treats each axis differently and, thus, can thus have two orientations. Often the orientation is easy to deduce from a combination of the given mappings and the types of positional scales in use. Thus, ggplot2 will by default try to guess which orientation the layer should have. Under rare circumstances, the orientation is ambiguous and guessing may fail. In that case the orientation can be specified directly using the \code{orientation} parameter, which can be either \code{"x"} or \code{"y"}. The value gives the axis that the geom should run along, \code{"x"} being the default orientation you would expect for the geom.
}

\section{Aesthetics}{

\code{geom_smooth()} understands the following aesthetics (required aesthetics are in bold):
\itemize{
\item \strong{\code{\link[=aes_position]{x}}}
\item \strong{\code{\link[=aes_position]{y}}}
\item \code{\link[=aes_colour_fill_alpha]{alpha}}
\item \code{\link[=aes_colour_fill_alpha]{colour}}
\item \code{\link[=aes_colour_fill_alpha]{fill}}
\item \code{\link[=aes_group_order]{group}}
\item \code{\link[=aes_linetype_size_shape]{linetype}}
\item \code{\link[=aes_linetype_size_shape]{linewidth}}
\item \code{weight}
\item \code{\link[=aes_position]{ymax}}
\item \code{\link[=aes_position]{ymin}}
}
Learn more about setting these aesthetics in \code{vignette("ggplot2-specs")}.
}

\section{Computed variables}{

These are calculated by the 'stat' part of layers and can be accessed with \link[=aes_eval]{delayed evaluation}. \code{stat_smooth()} provides the following variables, some of  which depend on the orientation:
\itemize{
\item \code{after_stat(y)} \emph{or} \code{after_stat(x)}\cr Predicted value.
\item \code{after_stat(ymin)} \emph{or} \code{after_stat(xmin)}\cr Lower pointwise confidence interval around the mean.
\item \code{after_stat(ymax)} \emph{or} \code{after_stat(xmax)}\cr Upper pointwise confidence interval around the mean.
\item \code{after_stat(se)}\cr Standard error.
}
}

\examples{
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth()

# If you need the fitting to be done along the y-axis set the orientation
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth(orientation = "y")

# Use span to control the "wiggliness" of the default loess smoother.
# The span is the fraction of points used to fit each local regression:
# small numbers make a wigglier curve, larger numbers make a smoother curve.
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth(span = 0.3)

# Instead of a loess smooth, you can use any other modelling function:
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth(method = lm, se = FALSE)

ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth(method = lm, formula = y ~ splines::bs(x, 3), se = FALSE)

# Smooths are automatically fit to each group (defined by categorical
# aesthetics or the group aesthetic) and for each facet.

ggplot(mpg, aes(displ, hwy, colour = class)) +
  geom_point() +
  geom_smooth(se = FALSE, method = lm)
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_smooth(span = 0.8) +
  facet_wrap(~drv)

\donttest{
binomial_smooth <- function(...) {
  geom_smooth(method = "glm", method.args = list(family = "binomial"), ...)
}
# To fit a logistic regression, you need to coerce the values to
# a numeric vector lying between 0 and 1.
ggplot(rpart::kyphosis, aes(Age, Kyphosis)) +
  geom_jitter(height = 0.05) +
  binomial_smooth()

ggplot(rpart::kyphosis, aes(Age, as.numeric(Kyphosis) - 1)) +
  geom_jitter(height = 0.05) +
  binomial_smooth()

ggplot(rpart::kyphosis, aes(Age, as.numeric(Kyphosis) - 1)) +
  geom_jitter(height = 0.05) +
  binomial_smooth(formula = y ~ splines::ns(x, 2))

# But in this case, it's probably better to fit the model yourself
# so you can exercise more control and see whether or not it's a good model.
}
}
\seealso{
See individual modelling functions for more details:
\code{\link[=lm]{lm()}} for linear smooths,
\code{\link[=glm]{glm()}} for generalised linear smooths, and
\code{\link[=loess]{loess()}} for local smooths.
}