In checking how well a multivariate calibration models data, an important graphic is the plot of the Y residuals, which is the difference between the true value and that predicted from the model. From this plot you can detect if there are samples that are outliers, and if there is any pattern in the residuals, which may indicate such things as time effects of measurements, or perhaps decay of sample or product over the course of the experimentation. To access the residuals plot from PLS overview, go to the menu and choose Plot->Residuals->General->Y-residuals vs. Predicted Y.

Calibration samples that have a high leverage or high x-residual value will not be detected by the residuals plot alone. The residual of such samples can be amplified by dividing the Y-residual by the standard deviation and the sample leverage. Using the studentized residuals which are scaled and corrected for leverage, makes the interpretation independent of the measurement units so that studentized residuals can be compared to each other on a common scale. In The Unscrambler® X it is easy to toggle between the residual plot and the studentized residual using the toolbar short cut keys: