grouper {nnDiag}R Documentation

Produce Ordered Groups of Data Elements

Description

The purpose of this function is to produce an object that can be used in many of the diagnostic tools from this package. It orders the reference set elements with respect to predictions and breaks them into groups with an arbitrary number of elements. Also included in the output is the number of times each element was used as a neighbor in the kNN classification and the residuals.

Usage

grouper(reference.set, predicted.set, nnIndex, group.size = 25, best = TRUE)

Arguments

reference.set vector of observed values
predicted.set vector of predicted values
nnIndex data.frame of nearest neighbors index
group.size a single integer, number of elements to be in each group
best logical indicating whether the function will use the exact group.size input or if it will find the “best” size nearest to the group.size input.

Details

The nnIndex matrix will have a column for each corresponding k used in the kNN classification. Each element in the matrix should be in reference to the reference.set vector position of the observed value.

If the remainder number of data elements do not fill a full group, the rest of that group will have NAs as place fillers. The best argument is to reduce the number of NAs in the last group. When best = TRUE the function will find the “best” group size, which is near the group.size input, that best fits the number of elements in the reference set. If best is set to FALSE it will use the group.size input exactly.

Value

Returns an object of class "nnDgrps", which is a list containing the following components:

ordered.data data frame of the inputed data ordered by predictions
reference.groups matrix of the groups of reference set elements where each column is a group.
predicted.groups matrix of the groups of predicted set elements where each column is a group.
residual.groups matrix of the groups of residuals where each column is a group.
group.size the number of data elements in each group.

Author(s)

Brian Walters walte137@msu.edu

References

McRoberts, R.E. (2009) Diagnostic tools for nearest neighbors techniques when used with satellite imagery, Remote Sensing of Environment. 113, 489–499.

See Also

Functions that use "nnDgrps" objects: scedast, outInflu, bias

Examples

data(LuceVolume)
data(LuceVolume_indx)

##Using the defaults
x <- grouper(LuceVolume$ref.volume, LuceVolume$pred.vol_k18, LuceVolume_indx)
x

##Not using the defaults
x <- grouper(LuceVolume$ref.volume, LuceVolume$pred.vol_k18, LuceVolume_indx,
group.size = 37, best = FALSE)
x

[Package nnDiag version 0.0-5 Index]