binning {sm}

Construct frequency table from raw data in 1, 2 or 3 dimensions.

Description

Given a vector, or a matrix with 1, 2 or 3 columns, this function constructs a frequency table associated with appropriate intervals covering the range of x.

Usage

binning(x, y, breaks, nbins)

Arguments

x

a vector or a matrix with either one, two or three columns, containing the original data.

y

a vector of data, for example response data, associated with the data in x.

breaks

either a vector or a matrix with two columns (depending on the dimension of x), assigning the division points of the axis, or the axes in the matrix case. It must not include Inf,-Inf or NAs, and it must span the whole range of the x points. If breaks is not given, it is computed by dividing the range of x into nbins intervals for each of the axes.

nbins

the number of intervals on each axis. If nbins is not supplied, a value is computed as round(log(n)/log(2) + 1).

Details

This function is called automatically (under the default settings) by some of the functions of the sm library when the sample size is large, to allow handling of datasets of essentially unlimited size. Specifically, it is used by sm.density, sm.regression, sm.ancova, sm.binomial and sm.poisson.

Value

In the vector case, a list is returned containing the following elements: a vector x of the midpoints of the bins excluding those with 0 frequecies, its associated matrix x.freq of frequencies, the co-ordinates of the midpoints, the division points, and the complete vector of observed frequencies freq.table (including the 0 frequencies), and the vector breaks of division points. In the matrix case, the returned value is a list with the following elements: a two-dimensional matrix x with the coordinates of the midpoints of the two-dimensional bins excluding those with 0 frequencies, its associated matrix x.freq of frequencies, the coordinates of the midpoints, the matrix breaks of division points, and the observed frequencies freq.table in full tabular form.

References

Bowman, A.W. and Azzalini, A. (1997). Applied Smoothing Techniques for Data Analysis: the Kernel Approach with S-Plus Illustrations. Oxford University Press, Oxford.

See Also

sm, sm.density, sm.regression, sm.binomial, sm.poisson, cut, table

Examples

# example of 1-d use
x  <- rnorm(1000)
xb <- binning(x)
xb <- binning(x, breaks=seq(-4,4,by=0.5))
# example of 2-d use
x <- rnorm(1000)
y <- 2*x + 0.5*rnorm(1000)
x <- cbind(x, y)
xb<- binning(x, nbins=12)