plspmSpatial

Set of functions to help in PLS-PM analysis in spatial ecological applications

The currently included functions are:

- plspmPredict
- plspm.groupsPredict
- plspmResiduals

@author: Javier Lopatin

plspmPredict:

This function predicts PLS-PM latent and measurement variables from a 'plspm' object ('plspm' R-package)

This is based on the publication: Shmueli, G., Ray, S., Estrada, J. M., & Chatla, S. (n.d.). The Elephant in the Room: Evaluating the Predictive Performance of Partial Least Squares (PLS) Path Models (2015). SSRN Electronic Journal SSRN Journal.

The script was adapted from the script: https://github.com/ISS-Analytics/pls-predict/blob/master/lib/PLSpredict.R

The adaptation was done in order to work directly with an plspm object.

WARNING: For the moment, only working with class(data) == data.frame object for prediction. Raster classes to be added

Usage:

plspmPred(pls, dat, ...)

Arguments:

• pls: An plspm object from the plspm package
• dat: data.frame or Raster Stack with the model predictors

Details:

The function plspmPredict estimates extrapolation values of Latent and Measurement Variables from and plspm object

Values:

An object of class plspmPredict is returned. The object returns a list with:

• mmData

  Matrix or RasterStack of the input measurement variables
  

• mmPredicted

  Matrix or RasterStack of the predicted all measurement variables
  

• mmResiduals

  Matrix of the residuals of all measurement variables. Only if validation data of the target endogenous variables are provided for validation
  

• Scores

  Matrix of the predicted Latent Variables scores [in ordination units]. Only if validation data of the target endogenous variables are provided for validation
  

• r_square

  Matrix of Squared Pearson's Correlation values of all measurement variables. Only if validation data of the target endogenous variables are provided for validation
  

• RMSE

  Matrix of Root-Mean-Square-Error values of all measurement variables Only if validation data of the target endogenous variables are provided for validation
  

• nRMSE

  Matrix of normalizedRoot-Mean-Square-Error [%] values of all measurement variables. Only if validation data of the target endogenous variables are provided for validation
  

• bias

  Matrix of bias values of all measurement variables. Only if validation data of the target endogenous variables are provided for validation
  

plspm.groupsPredict:

Usage:

plspm.groupsPredict(pls, pls.groups, train.groups, dat)

Details:

This function has the same functions as plspmPredict, but uses a plspm.groups object as input. Therefore, it gives a list of predicted scores, measurement variables, and residuals for the 'general', 'group1', and 'group2' models.

plspmRsiduals:

Usage:

plspmRsiduals(pls)

Arguments:

• pls: An plspm object from the plspm package
• dat: data.frame or Raster Stack with the model predictors

Details:

This function obtain residuals for all Latent and Measurement variables from an 'plspm' object

Values:

An object of class plspmResiduals is returned. The object returns a list with:

• inner_residuals

  Matrix of residual values for the Latent Variables
  

• outer_residuals

  Matrix of residual values for the Measurement Variables