Calibration Model Wizard
- Step 6 - Factor Analysis
The Factor Analysis is part of all calibration models except the MLR-Model. When using the MLR-Model
this step will be omitted.
The PLS calibration model analyses the data and divides variables into
principle components. They are also called factors forming an n-dimensional
factor space. The number of factors taken into account for calculation
of a calibration model is very important, because if the user selects
too many, he will interpret spectral noise. This reduces the quality of
the calibration model significantly.
Thus factor selection is very important and the predictive residual
error sum of squares (PRESS) plot provides a good overview on how many
factors need to be used for calculation. It gives an indication of the
model error vs. the number of factors.
Within the wizard the software proposes a number of analyzed factors.
This number can be accepted or may be changed interactively later by moving
the vertical line or increasing / decreasing the values using the spin
The PRESS plot shows a graph of the PRESS values plotted against the
factors and will always look like a decaying curve. In the figure above,
the total number of factors is 10 according to the number of principle
components. The PRESS values indicate a high priority for factors 1 to
3. Factors beyond 3 do not contain any additional information useful for
calibration. The vertical green line indicates the automatically proposed
number of factors for which the model reached its minimum. In the present
case, the number of factors at the minimum is three. The user can accept
this number or change it now or later on.
Just click the Next
> button to proceed to the next step.
Clicking the Cancel button will
abort creating a new calibration.