[PLS-recalculations][PLS-predictions][R2 values][Q2 values][PCA-scores][PCA-scores ext. pred.][Weights][PLS-loadings][PLS-partial weights][PLS-coefficients][Selected variables]
The List menu displays some of the numerical data computed by GOLPE during the building and validation of PCA and PLS models.
List>>>PCA-scores ext. pred.
These commands produce the display of the following data on the main window:
Lists the values of the Y-variable(s) recalculated by the model for each object and for each component.
Provided that the model has already been validated, this command lists the values of the Y-variable(s) predicted by the model for each object and for each component. Please note that the values shown correspond to predictions obtained from reduced models used in the validation, and therefore they will be different from the recalculated values.
Lists for each Y-variable the SDEC values and the corresponding r2 values, component by component.
Provided that the model has already been validated, this command lists for each Y-variable the SDEP values and the corresponding q2 values, component by component.
Lists the values of the scores for all the objects in the dataset and all model dimensions.
PCA-scores ext. pred.
This option is active only when the User made external PCA predictions using the command Utilities>>>PCA-predictions. Please consult this command to learn more about PCA-predictions.
Lists the values of the scores for all the objects in the external dataset and all model dimensions.
All this three commands present data for each X-variable in the data file. Very often the data file contains a lot of X-variables and it is preferable to display only the most significative values. In order to filter the smaller values the User is first prompted to enter a cutoff value:
Absolute values smaller than the cutoff will not be listed, but if the User enters 0.000 all variables will be listed. The default cutoff is 0.01.
Lists the weights applied to the variables. Usually GOLPE applies a weight of 1.000 to every variable in the data file and so, all the variables have an equal importance in the modeling. The exceptions are:
When the User applying autoscaling, BUW or change the weight factor with the commands Pretratment>>>Classic Pretreatment>>>Set-up pretreatment, the weights of the variables will change accordingly.
Lists the loading values obtained in the PLS modeling. See Background section for a definition of loadings in the context of PLS.
Lists the PLS partial-weights W's from where the loadings are calculated. See Background section for a definition of partial-weights.
From a PLS model, when there is only one Y-variable, it is possible to build a pseudo-MLR model, in which a single coefficient (pseudo-coefficients) multiplies each X-variable. These coefficients express the effect of each single X-variable on the dependent variable and can be useful to interpret the whole model in some contexts. It is important to notice that pseudo-coefficients do vary with the PLS model dimensionality.
There are two ways to present the PLS coefficients:
They can be directly used for future predictions on raw X-data obtained for new compounds.
They correspond to the real coefficients multiplied by the X weights. When the X-matrix has been autoscaled, they are needed to compare the relative importance of individual X-variables in the pseudo-MLR model. These coefficients are therefore more reliable for interpretation.
All this two commands present data for each X-variable in the data file. Very often the data file contains a lot of X-variables and it is preferable to display only the most significative values. In order to filter the smaller values the User is first prompted to enter a cutoff value:
Absolute values smaller than the cutoff will not be listed, but if the User enters 0.000, all variables will be listed. The default cutoff is 0.01.
For each dimensionality of the PLS model, the real pseudo-coefficients for each X-variable, are listed.
For each dimensionality of the PLS model, the weighted pseudo-coefficients for each X-variable, are listed.
Lists the indexes (from the original data file) of the variables selected in a previously made D-optimal preselection or F. Factorial variable selection procedures.
List>>>Selected variables>>>D-optimal selection
List>>>Selected variables>>>F. Factorial selection
IMPORTANT: the D-optimal selection list or the F. Factorial selection list will appear only after performing a variable selection procedure
A dialog window like this is presented:
D-optimal selection list
F. Factorial selection list
This list contains all the previous variable selection procedures performed on this data file. Each procedure is identified by a sequential number, the initial number of variables, the final number of variables, the number of components and the hour and data when the procedure was finished. Click on any item to select it. The selected items will be included in the Selection input field.
The selection from the above list is shown in the Selection input field.
When the OK button is pressed, all the variables selected in the chosen procedure will be listed on the main window.