plot.sPLSDA.performance

Plots the cross-validated classification error of a sPLS-DA as a function of the number of components.

plot.sPLSDA.performance(
  DEprot.sPLSDA.object,
  metric = "both",
  distance = "all",
  line.colors = NULL,
  title = "**Cross-validated classification error**"
)

Arguments

DEprot.sPLSDA.object

An object of class DEprot.sPLSDA.

metric

String indicating which error to display. One among: 'both', 'overall', 'BER' (balanced error rate). Default: "both".

distance

String indicating the prediction distance(s) to display. One among: 'all', 'max.dist', 'centroids.dist', 'mahalanobis.dist'. Default: "all".

line.colors

Color-vector used for the distances. Default: NULL (automatic colors).

title

String indicating the title of the plot (markdown annotation supported). Default: "**Cross-validated classification error**".

Value

A ggplot object.

Details

The balanced error rate is the average of the error rates of each class, and it is the value to read on an unbalanced design: with 9 samples in one group and 3 in the other, a model that assigns everything to the larger group already reaches an overall error of 25

The dashed gray line marks the error of a random assignment, that is one minus the frequency of the largest class. A model sitting on that line is not discriminating anything, however clean its scatter plot looks.

See also

perform.sPLSDA, plot.sPLSDA.tuning, plot.sPLSDA.auroc

Author

Sebastian Gregoricchio

Examples

# \donttest{
# the performances are estimated by resampling the samples: with 12 samples in 3 groups
# the leave-one-out scheme is the only one leaving enough samples in each fold
splsda <- perform.sPLSDA(DEprot.object = DEprot::test.toolbox$dpo.imp,
                         group.column = "condition",
                         keepX = 5,
                         validate = TRUE,
                         validation = "loo")
#> Estimating the classification performances by cross-validation...

plot.sPLSDA.performance(DEprot.sPLSDA.object = splsda)

# }