PLS Results - Advanced Tab

Partial Least Squares (PLS)

Select the Advanced tab of the PLS Results dialog box to access the options described here.

Include test samples. Select this check box to include the test sample in the spreadsheets of predictions, residuals, and scores. Cases from the test sample will be shown in red.

Element Name	Description
Prediction	The following options are available in the Prediction group box.
Use original scale	Select this check box to generate predictions and residuals on the scale of the original or raw data. If not selected, predictions and residuals will be based on the normalized/scaled data.
Predictions	Click this button to produce a spreadsheet of predictions.
Residuals	Click this button to produce a spreadsheet of residuals. Residuals are defined as the difference between the original variables and the predictions of the PLS model. In other words, residuals are the unmodeled parts of the data that could not be matched by the predictions of the model. Large residuals are indications of abnormality in the data that cannot be sufficiently predicted by the model. The ability to detect outliers is a useful feature of PLS that can be utilized for process monitoring (see MSPC Technical Notes) and quality control.
X Scores (t)	Click this button to produce a spreadsheet of t-scores. Scores are the representation of the original X variables in the new coordinate system, i.e., the system of the principal components.
Save X Scores	Click this button to display a standard variable selection dialog box, which is used to select variable(s) to be displayed together with the t-scores [see the description of the X Scores (t) option, above]. After you select the variable(s), a spreadsheet containing the specified variable(s) will be produced in an individual window (regardless of the settings on the Options dialog box - Output Manager tab or the Analysis/Graph Output Manager dialog box). You can, however, add the spreadsheet to a workbook or report using the or buttons, respectively. Note that in order to save the spreadsheet, you must select the spreadsheet and select Save or Save As from the File menu. This is useful if you want to use the residual values for further study with other STATISTICA analyses.
Y Scores (u)	Click this button to produce a spreadsheet of y-scores. Scores are the representation of the original Y variables in the new coordinate system, i.e., the system of the principal components.
Save Y Scores	Click this button to display a standard variable selection dialog box, which is used to select variable(s) to be displayed together with the y-scores [see the description of the Y Scores (u) option, above]. After you select the variable(s), a spreadsheet containing the specified variable(s) will be produced in an individual window (regardless of the settings on the Options dialog box - Output Manager tab or the Analysis/Graph Output Manager dialog box). You can, however, add the spreadsheet to a workbook or report using the or buttons, respectively. Note that in order to save the spreadsheet, you must select the spreadsheet and select Save or Save As from the File menu. This is useful if you want to use the residual values for further study with other STATISTICA analyses.
Coefficients	Click this button to generate a spreadsheet of the PLS coefficients. The coefficients of a PLS model relate the u-scores of the dependent variable to the t-scores of the independent (continuous and/or categorical).
X Weights (w)	Click this button to produce a spreadsheet containing the weights of the X variables.
X Loadings (p)	Click this button to create the matrix of the p-loading factors for the principal components in spreadsheet format. The loading factors determine the orientation of the principal component axes with respect to the original coordinate system (defined by X). Loading factors are used to analyze the influence of the original variables in determining the PLS model.
Y Loadings (q)	Click this button to create the matrix of the p-loading factors in spreadsheet format for the PLS components.
Eigenvalues	Click this button to produce a spreadsheet of the vector of eigenvalues of the principal components.
Eigenvalues	Use the options in this group box to generate line plots for a specified number of principal eigenvalues.
Scree plot	Click this button to create an eigenvalue scree plot  (Cattell, 1966) for the extracted principal components. By default, only the extracted eigenvalues are included, but you can extend this number (up to the maximum number of the eigenvalues) using Number of eigenvalues option below.
Number of eigenvalues	In this box, specify how many eigenvalues to be included in the scree plot.
D-To-Model (X)	Click this button to produce a spreadsheet of distance-to-model for the X observations in the data set. Distance-to-model plays an important role in process control since it measures the squared perpendicular distance of an observation from the normal plane. Distance-to-model is used as an indication of whether a new case is within the domain of normality. Hence, it can be used for detecting outliers.
D-To-Model (X)	Click this button to produce the D-To-Model (X) in line plot format.
D-To-Model (X)	Click this button to produce the D-To-Model (X) in histogram format.
D-To-Model (Y)	Click this button to produce a spreadsheet of distance-to-model for the Y observations in the data set. Distance-to-model plays an important role in process control since it measures the squared perpendicular distance of an observation from the normal plane. Distance-to-model is used as an indication of whether a new case is within the domain of normality.  Hence, it can be used for detecting outliers.
D-To-Model (Y)	Click this button to produce the D-To-Model (Y) in line plot format.
D-To-Model (Y)	Click this button to produce the D-To-Model (Y) in line histogram format.
Descriptives	Click this button to produce a spreadsheet of various statistics of the original variables such as number of valid cases, means, standard deviations, and scale.
Scaled data	Click this button to generate a spreadsheet of the pre-processed variables X and Y. Pre-processing involves the application of a linear transformation which transforms the original data set to a new set of variables with zero mean and unit (or user specified) standard deviation.