predict.mat {analogue}

Predict method for Modern Analogue Technique models

Description

Predicted values based on a MAT model object.

Usage

## S3 method for class 'mat'
predict(object, newdata, k, weighted = FALSE,
        bootstrap = FALSE, n.boot = 1000,
        probs = c(0.01, 0.025, 0.05, 0.1), ...)

Arguments

object

an object of mat.

newdata

data frame; required only if predictions for some new data are required. Mst have the same number of columns, in same order, as x in mat. See example below or join for how to do this. If newdata not provided, the fitted values are returned.

k

number of analogues to use. If missing, k is chosen automatically as the k that achieves lowest RMSE.

weighted

logical; should the analysis use the weighted mean of environmental data of analogues as predicted values?

bootstrap

logical; should bootstrap derived estimates and sample specific errors be calculated-ignored if newdata is missing.

n.boot

numeric; the number of bootstrap samples to take.

probs

numeric; vector of probabilities with values in [0,1].

...

arguments passed to of from other methods.

Details

This function returns one or more of three sets of results depending on the supplied arguments:

Fitted values:

the fitted values of the mat model are returned if newdata is missing.

Predicted values:

the predicted values for some new samples are returned if newdata is supplied. Summary model statistics and estimated values for the training set are also returned.

Bootstrap predictions and standard errors:

if newdata is supplied and bootstrap = TRUE, the predicted values for newdata plus bootstrap estimates and standard errors for the new samples and the training set are returned.

The latter is simply a wrapper for bootstrap(model, newdata, ...) - see bootstrap.mat.

Value

A object of class predict.mat is returned if newdata is supplied, otherwise an object of fitted.mat is returned. If bootstrap = FALSE then not all components will be returned.

observed

vector of observed environmental values.

model

a list containing the model or non-bootstrapped estimates for the training set. With the following components:

• estimatedestimated values for "y", the environment.

• residualsmodel residuals.

• r.squaredModel R^2 between observed and estimated values of "y".

• avg.biasAverage bias of the model residuals.

• max.biasMaximum bias of the model residuals.

• rmsepModel error (RMSEP).

• knumeric; indicating the size of model used in estimates and predictions.

bootstrap

a list containing the bootstrap estimates for the training set. With the following components:

• estimatedBootstrap estimates for "y".

• residualsBootstrap residuals for "y".

• r.squaredBootstrap derived R^2 between observed and estimated values of "y".

• avg.biasAverage bias of the bootstrap derived model residuals.

• max.biasMaximum bias of the bootstrap derived model residuals.

• rmsepBootstrap derived RMSEP for the model.

• s1Bootstrap derived S1 error component for the model.

• s2Bootstrap derived S2 error component for the model.

• knumeric; indicating the size of model used in estimates and predictions.

sample.errors

a list containing the bootstrap-derived sample specific errors for the training set. With the following components:

• rmsepBootstrap derived RMSEP for the training set samples.

• s1Bootstrap derived S1 error component for training set samples.

• s2Bootstrap derived S2 error component for training set samples.

weighted

logical; whether the weighted mean was used instead of the mean of the environment for k-closest analogues.

auto

logical; whether "k" was choosen automatically or user-selected.

n.boot

numeric; the number of bootstrap samples taken.

predictions

a list containing the model and bootstrap-derived estimates for the new data, with the following components:

• observedthe observed values for the new samples — only if newenv is provided.

• modela list containing the model or non-bootstrapped estimates for the new samples. A list with the same components as model, above.

• bootstrapa list containing the bootstrap estimates for the new samples, with some or all of the same components as bootstrap, above.

• sample.errorsa list containing the bootstrap-derived sample specific errors for the new samples, with some or all of the same components as sample.errors, above.

method

the dissimilarity measure used to fit the underlying MAT models.

quantiles

probability quantiles of the pairwise dissimilarities computed from the training set.

minDC

smallest distances between each sample in newdata and the training set samples.

Dij

dissimilarities between newdata and training set samples.

Author(s)

Gavin L. Simpson

References

Birks, H.J.B., Line, J.M., Juggins, S., Stevenson, A.C. and ter Braak, C.J.F. (1990). Diatoms and pH reconstruction. Philosophical Transactions of the Royal Society of London; Series B, 327; 263–278.

See Also

mat, bootstrap.mat

Examples

## Imbrie and Kipp example
## load the example data
data(ImbrieKipp)
data(SumSST)
data(V12.122)

## merge training and test set on columns
dat <- join(ImbrieKipp, V12.122, verbose = TRUE)

## extract the merged data sets and convert to proportions
ImbrieKipp <- dat[[1]] / 100
V12.122 <- dat[[2]] / 100

## fit the MAT model using the chord distance measure
(ik.mat <- mat(ImbrieKipp, SumSST, method = "chord"))

## predict for V12.122 data
predict(ik.mat, V12.122)