plot_classifX_indepPred: Plot Independent Predictions

In bpollner/aquap2: Multivariate Data Analysis Tools for R including Aquaphotomics Methods

Plot Independent Predictions

Description

Calculate and plot predictions from independent, manually provided data. One or more of the implemented classifier-models (see e.g. calc_discrimAnalysis_args and the links to other classifier functions there) have to be present in the cube. The manually provided data in indepDataset are then projected into each single classification model present in the cube, and the results are validated using either the class variable present in the independent dataset that has the exactly same name as the class variable used to generate the models, or a user-defined class variable (parameter icv) can be used for validation.

Usage

plot_classifX_indepPred(
  indepDataset,
  cube,
  ccv = NULL,
  icv = NULL,
  pl = TRUE,
  toxls = "def",
  info = "def",
  confirm = "def",
  predList = NULL,
  apPlot = "def",
  ...
)

Arguments

indepDataset	The dataset containing the independent data. An object of class 'aquap_data' as produced by gfd.
cube	An object of class 'aquap_cube' as produced by gdmm. It is an error to have no classification models in the cube.
ccv	"Cube class variable", character vector or NULL. The names of one or more class variables in the cube on which classification models have been calculated. Leave at the default NULL to use all of the class variables on which a classification model has been calculated, or provide a character vector with valid variable names for a sub-selection. For the selected variables, predictions from the data in the independent dataset will be made. If argument icv is left at its default NULL, class variables with exactly the same name are looked for in the independent dataset and, if present, are used for validating the predictions.
icv	"Independent class variable", character vector or NULL. The names of class variables in the independent dataset. If left at the default NULL, class variables in the independent dataset with exactly the same name(s) as specified in argument ccv are looked for and, if present, are used for validating the predictions. If a character vector is provided, it has to have the same length as the one in ccv, and those variables will be used, in the given sequence, for validating the predictions.
pl	Logical, defaults to TRUE. If predicted data should be plotted at all. If FALSE, only the calculation and the (possible) export to an excel file (see details) will be performed.
toxls	'def' or logical. If left at the default 'def' the value from cl_indepPred_exportToExcel in the settings file is used. Set to TRUE or FALSE to directly control if export of predicted data to excel should be performed or not.
info	'def' or logical. If left at the default 'def' the value from cl_indepPred_showInfo in the settings file is used. Set to TRUE or FALSE to directly control if information regarding the pairing of class variables in the model and those in the independent dataset used for validation should be displayed.
confirm	'def' or logical. If left at the default 'def' the value from cl_indepPred_confirm in the settings file is used. Set to TRUE or FALSE to directly control if manual confirmation is required after the (possible) display of pairing-information (see above). Ignored if info is FALSE.
predList	NULL or list. If left at the default null, the independent dataset is used to make predictions on all available classification models, resulting in a list containing the predictions. If this list is in turn provided to predList, no calculations are performed and the results are plotted straight away.
apPlot	The analysis procedure used for plotting.
...	General plotting parameters, see XXX.

Details

For every single element in the cube, i.e. for every split-variation of the original dataset (as produced in gdmm the according subgroups within the independent dataset are constructed. In case of a dataset resulting in no observations the process is aborted. Also, the data pre- and post- treatments (see dpt_modules) as defined in the analysis procedure used to produce the cube are applied to the independent dataset resp. to its subgroups as defined by the application of possible split-variables (see above). If toXls is TRUE, the results of the predictions will be exported to Excel.

Value

An (invisible) list containing the numerical results of the predictions, and if parameter toXls is TRUE, these data are exported to an excel file in the results folder as well.

See Also

Other Classification functions: calc_NNET_args, calc_SVM_args, calc_discrimAnalysis_args, calc_randomForest_args

Other Plot functions: plot,aquap_cube,missing-method, plot,aquap_data,missing-method, plot_aqg(), plot_da,aquap_cube-method, plot_nnet,aquap_cube-method, plot_pca,aquap_cube-method, plot_pls,aquap_cube-method, plot_pls_indepPred(), plot_rnf,aquap_cube-method, plot_simca,aquap_cube-method, plot_svm,aquap_cube-method

Examples

## Not run: 
fd <- gfd() # loading or importing the rawdata
fd1 <- ssc(fd, C_Foo!="bar") # manually splitting up the dataset
fd2 <- ssc(fd, C_Foo=="bar") 
cube <- gdmm(fd1) # this is assuming that the standard analysis procedure is set
# up to perform a classifier method
# we are using `fd1` to produce the cube, and then `fd2` as independent dataset
# to perform independent predictions on all the models in the cube
predictions <- plot_classifX_indepPred(fd2, cube)
predictions <-  plot_classifX_indepPred(fd2, cube, icv="C_blabla", pl=FALSE)
# to redirect the original validation to class-variable "C_blabla"

## End(Not run)

bpollner/aquap2 documentation built on March 29, 2024, 7:33 a.m.