Understanding PLS Graphs
main topic 

PLS graph

Definition

Use to...

Model selection plot

Scatterplot of the R2 and predicted R2 values as a function of the number of components. The vertical line indicates number of components in the optimal model.

Compare the modeling and predictive powers of models with different numbers of components.

Response plot

Scatterplot of the fitted and cross-validated responses versus the actual responses.

Show how well the model fits and predicts. Large differences in fitted and cross-validated values indicate leverage points.

Coefficient plot

Projected scatterplot of the unstandardized regression coefficients.

View sign and the magnitude of the relationship between predictors and responses.

Standardized coefficient plot

Projected scatterplot of the standardized regression coefficients.

View sign and the magnitude of the relationship between predictors and responses when predictors are not on the same scale.

Distance plot

Scatterplot of each observation's distance from the x-model and distance from y-model.

Identify leverage points and outliers.

Residual histogram

Histogram of the standardized residuals.

Check the normality of your residuals. Histograms should show a bell-shaped distribution.

Residual normal probability plot

Scatterplot of the standardized residuals versus the normal scores.

Check the normality of your residuals. Points should follow a straight line.

Residual versus fit plot

Scatterplot of the standardized residuals versus the fitted responses.

Identify outliers and check for patterns in the residuals.

Residual versus leverage plot

Scatterplot of the standardized residuals versus leverages.

Identify outliers and leverage points simultaneously.

Residual fourpack

Layout of residual histogram, residual normal plot, residual versus fit plot, and residual versus order plot on one page.

View residual plots simultaneously.

Score plot

Scatterplot of the x-scores from the first and second components.

Display the overall configuration of the data using the first two components to identify leverage points or clusters of points.

3D score plot

3D scatterplot of the x-scores from the first, second, and third components.

Display the overall configuration of the data using the first three components to identify leverage points or clusters of points.

Loading plot

Connected scatterplot of the x-loadings from the first and second components.

Display the correlation between the loadings of each predictor on the first and second components. Compare the importance of predictors to the model.

Residual X plot

Connected scatterplot of the x-residuals, in which each line represents an observation and has as many points as predictors.

Identify observations or predictors that are poorly explained by the model.

Calculated X plot

Connected scatterplot of the x-calculated values, in which each line represents an observation and has as many points as predictors.

Identify observations or predictors that are poorly explained by the model.