% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/gridVectors.R
\name{gridVectors}
\alias{gridVectors}
\alias{gridVectors,ANY-method}
\alias{gridVectors,SingleCellExperiment-method}
\title{Summarize vectors into a grid}
\usage{
gridVectors(x, embedded, ...)

\S4method{gridVectors}{ANY}(
x,
embedded,
resolution = 40,
scale = TRUE,
as.data.frame = TRUE,
return.intermediates = FALSE
)

\S4method{gridVectors}{SingleCellExperiment}(x, embedded, ..., use.dimred = 1)
}
\arguments{
\item{x}{A numeric matrix of low-dimensional coordinates, e.g., after t-SNE.

\item{embedded}{A low-dimensional projection of the velocity vectors into the embedding of \code{x}.
This should be of the same dimensions as \code{x} and is typically produced by \code{\link{embedVelocity}}.}

\item{...}{For the generic, further arguments to pass to specific methods.

For the SingleCellExperiment method, further arguments to pass to the ANY method.}

\item{resolution}{Integer scalar specifying the resolution of the grid,
in terms of the number of grid intervals along each axis.}

\item{scale}{Logical scalar indicating whether the averaged vectors should be scaled by the grid resolution.}

\item{as.data.frame}{Logical scalar indicating whether the output should be a data.frame.
If \code{FALSE}, a list of two matrices is returned.}

\item{return.intermediates}{Logical scalar indicating whether intermediate objects should also be returned.
This enforces \code{as.data.frame=FALSE} and throws a warning if it is \code{TRUE}.}

\item{use.dimred}{String or integer scalar specifying the reduced dimensions to retrieve from \code{x}.}
}
\value{
If \code{as.data.frame=FALSE}, a list is returned containing \code{start} and \code{end},
two numeric matrices with one row per non-empty block in the grid and one column per column in \code{x}.
\code{start} contains the mean location of all cells inside that block,
and \code{end} contains the endpoint after adding the (scaled) average of the block's cell's velocity vectors.

If \code{as.data.frame=TRUE}, a data.frame is returned with numeric columns of the same contents as the list above.
Column names are prefixed by \code{start.*} and \code{end.*}.

If \code{return.intermediates=TRUE}, a list is returned (irrespective of the value of \code{as.data.frame})
that in addition to \code{start} and \code{end} also contains intermediate objects \code{limits}
(the ranges in x and y), \code{delta} (the grid intervals in x and y), \code{categories} (a
DataFrame with integer row and column indices for each cell that specify the grid field that it is
contained in), \code{grp} (numerical index of grid fields for each cell) and \code{vec} (velocity
vectors for non-empty grid fields).
}
\description{
Summarize the velocity vectors into a grid, usually for easy plotting.
}
\details{
This partitions the bounding box of \code{x} into a grid with \code{resolution} units in each dimension.
The locations and vectors of all cells in each block are averaged to obtain a representative of that block.
This is most obviously useful for visualization to avoid overplotting of velocity vectors.

If \code{scale=TRUE}, per-block vectors are scaled so that the median vector length is comparable to the spacing between blocks.
This improves visualization when the scales of \code{x} and \code{embedded} are not immediately comparable.
}
\examples{
tsne.results <- matrix(rnorm(10000), ncol=2)
tsne.vectors <- matrix(rnorm(10000), ncol=2)

out <- gridVectors(tsne.results, tsne.vectors)

# Demonstration for plotting.
plot(tsne.results[,1], tsne.results[,2], col='grey')
arrows(out$start.1, out$start.2, out$end.1, out$end.2, length=0.05)

}
\seealso{