mdatools (version 0.9.1)

simcamres: Results of SIMCA multiclass classification

Description

simcamres is used to store results for SIMCA multiclass classification.

Usage

simcamres(cres, pred.res)

Arguments

cres

results of classification (class classres).

pred.res

prediction results from each model (as pcares)

Value

Returns an object (list) of class simcamres with the same fields as classres plus extra fields for Q and T2 values and limits:

c.pred

predicted class values.

c.ref

reference (true) class values if provided.

T2

matrix with T2 values for each object and class.

Q

matrix with Q values for each object and class.

T2lim

vector with T2 statistical limits for each class.

Qlim

vector with Q statistical limits for each class.

The following fields are available only if reference values were provided.

tp

number of true positives.

fp

nmber of false positives.

fn

number of false negatives.

specificity

specificity of predictions.

sensitivity

sensitivity of predictions.

Details

Class simcamres inherits all properties and methods of class classres, plus store values necessary to visualise prediction decisions (e.g. Cooman's plot or Residuals plot).

In cotrast to simcares here only values for optimal (selected) number of components in each individual SIMCA models are presented.

There is no need to create a simcamres object manually, it is created automatically when make a SIMCAM model (see simcam) or apply the model to a new data (see predict.simcam). The object can be used to show summary and plots for the results.

See Also

Methods for simcamres objects:

print.simcamres	shows information about the object.
summary.simcamres	shows statistics for results of classification.
plotResiduals.simcamres	makes Q vs. T2 residuals plot.
plotCooman.simcamres	makes Cooman's plot.

Methods, inherited from classres class:

showPredictions.classres	show table with predicted values.
plotPredictions.classres	makes plot with predicted values.

Check also simcam.

Examples

# NOT RUN {
## make a multiclass SIMCA model for Iris data and apply to test set
library(mdatools)

# split data 
caldata = iris[seq(1, nrow(iris), 2), 1:4]
se = caldata[1:25, ]
ve = caldata[26:50, ]
vi = caldata[51:75, ]

testdata = iris[seq(2, nrow(iris), 2), 1:4]
testdata.cref = iris[seq(2, nrow(iris), 2), 5]

# create individual models
semodel = simca(se, classname = 'setosa')
semodel = selectCompNum(semodel, 1)

vimodel = simca(vi, classname = 'virginica')
vimodel = selectCompNum(vimodel, 1)

vemodel = simca(ve, classname = 'versicolor')
vemodel = selectCompNum(vemodel, 1)

# combine models into SIMCAM object, show statistics 
model = simcam(list(semodel, vimodel, vemodel), info = 'Iris data')
res = predict(model, testdata, testdata.cref)
summary(res)

# show predicted values
showPredictions(res)

# plot predictions
par(mfrow = c(2, 2))
plotPredictions(res)
plotPredictions(res, nc = 1)
plotPredictions(res, nc = c(1, 2))
plotPredictions(res, show.labels = TRUE)
par(mfrow = c(1, 1))

# show residuals and Cooman's plot

par(mfrow = c(2, 2))
plotCooman(res)
plotCooman(res, nc = c(1, 3))
plotResiduals(res)
plotResiduals(res, nc = 3)
par(mfrow = c(1, 1))

# }