PLS Results - Plots Tab

Partial Least Squares (PLS)

Select the Plots tab of the PLS Results dialog box to access options for data diagnostics (whether on a variable or a case wise basis).

Element Name Description
Contribution The Contributions group box contains options to calculate variable contributions to the scores.
Contribution plot Click this button to generate a histogram of variable contributions to the score of the observation specified in the Obs. box, in the direction of the principal component selected in the First list (see First option description below). Different analysis variables make different contributions to the scores. For more details, see PCA and PLS Technical Notes.
Click this button to generate the same information described above in spreadsheet format.
Obs In this box, specify the observation for which you want to generate a histogram or spreadsheet of variables contributions (see above).
Lineplot (p/q) Click this button to create a line plot of the loading factors for the principal component selected in the First list (see the First option description below). For the p-loadings, select the X(t, p) option button in the First group box, otherwise select Y(u, q) for the q-loadings. Then select (highlight) a principal component in the First list for which you want to generate the line plot. You can generate more than one graph by making multiple selections in the First list.
Scatterplot (p/q) Click this button to create a loading scatterplot for the principal components selected in the First and Second components lists. First, specify a subject for the x-axis by selecting either the X(t, p) or Y(u, q) option button in the First group box. Then, select (highlight) one of the components in the components list (also in the First group box). Repeat these steps to select a subject for the y-axis using the options in the Second group box, and click the Scatter (p/q) button to create the graph. For example, select X(t, u) and Component 1 in the First group box, and Y(u, q) and Component 2 in the Second to create a scatterplot of the x-loadings for Component 1 against the y-loadings of Component 2.
Note: you can generate more than one graph by making multiple selections in the First and Second lists. For example, you can create the scatterplots of Components 1 and 2 vs. 3 if you select Components 1 and 2 in the First list and 3 in the Second.

The loading scatterplots can be used to examine the relationship between the process variables and the type of influence they exert on the PC model. Clustered variables are positively related to each other and negatively related to those located on the opposite quadrant. Clustered variables also influence the PC model in similar ways (see PCA and PLS Technical Details).

Lineplot (t/u) Click this button to create a line plot of the scores along the direction of the principal component specified in the First list. For the t-scores, select the X(t, p) option button, otherwise select Y(u, q) for the u-scores. Then select (highlight) a principal component from the First list for which you want to generate the line plot. You can generate more than one graph by making multiple selections from the First list.
Click this button to display the data generated in the Lineplot (t/u) option (described above) in spreadsheet format.
Scatter (t/u) Click this button to create a scores scatterplot for the principal components selected in the First and Second components lists. First, select a subject for the x-axis by selecting either the X(t, p) or Y(u, q) option button in the First group box. Then, select (highlight) one of the displayed components in the components list (also in the First group box). Repeat the above steps to select a subject for the y-axis using the options in the Second group box, and click the Scatter (t/u) button to create the graph. For example, select X(t, u) and Component 1 in the First list and Y(u, q) and Component 2 in the Second to create a scatterplot of the x-scores for Component 1 against the y-scores of Component 2. This graph will also display an ellipse, defined as
Observations with score values falling outside this ellipse may be outliers.
Note: you can generate more than one graph by making multiple selections in the First and Second lists. For example, you can generate the scatterplots of Components 1 and 2 vs. 3 if you select Components 1 and 2 in the First list and 3 in the Second.
Click this button to produce the data generated in the Scatter (t/u) option (described above) in spreadsheet format.
Biplot Select this check box to augment the component scores scatterplot with variable projections. The specific variables projected into the component space can be selected by clicking the Bi. Vars button.
Standardized plot Select this check box to standardize the biplot, that is, display the standardized component scores and standardized variable projections.
First The options in this group box will determine the type of data you want displayed in graphs of line and scatter plots (see above).
X(t, p) Select this option button to assign the t-scores/p-loadings as the y-axis of the lineplots or as the x-axis of scatter plots.
Y(u, q) Select the option button to assign the u-scores/q-loadings as the y-axis of the line plots or as the x-axis for scatter plots.

The Components list displays the list of principal components for the PLS model.

Second The options in this group box will determine the type of data you want displayed in scatter plots (see above).
X(t, p) Select this option button to assign the t-scores/p-loadings as the y-axis of the scatter plots.
Y(u, q) Select this option button to assign the u-scores/q-loadings as the y-axis of the scatterplots.

The Components list displays the list of principal components for the PLS model.

Plot labels The following three options are available in the Plot labels group box.
No name/number Select this option button to exclude variable names or case numbers (whichever is appropriate) from the graphs of scores and loading factors (lines and scatter plots).
Var/case number Select this option button to include variable or case numbers (whichever is appropriate) in the graphs of scores and loading factors (lines and scatter plots).
Var/case name Select this option button to include variable or case names (whichever is appropriate) in the graphs of scores and loading factors (lines and scatter plots).
Limits (t/u) Use the options in this group box to define limits for the score plots (line or scatter).
Multiply of std. dev. Select this option button to define the upper and lower bounds for the score limits as  control  standard deviation.
Compute from p If you select this option button, STATISTICA will assume that the scores are normally distributed and the 99% confidence levels are used to define the upper and lower limits.
Control Specify the control limits. When the Multiply std. dev. option button is selected (see above), this value is multiplied by the computed standard deviation for defining the upper and lower limits.
Warning Select this check box to include an upper and a lower warning limit in the scores graph. Specify the warning limit in the adjacent box. For the Multiply std. dev. option, this value will be multiplied by the standard deviation for the scores to define the lower and upper warning limits.