Metrohm Vision Manual - Manual (Page 133)

▪▪▪▪▪▪▪ 131
4. Click on OK.
5.4
Regression
Regression program allows users to develop a calibration equation from the calibration set. Vision
offers two types of calibration development: Multilinear Regression and Partial Least Squares.
5.4.1 Multilinear Regression
Multilinear Regression (MLR) is a least squares method that uses spectral information at one or
several wavelengths. In the simplest case (one wavelength), the method reduces to simple Linear
Regression.
Preparation
1. Click on Mode/Quantitative Analysis/Regression to enter the calibration development
program. The last active project will be opened. Expand the product for which you want to
create the equation. The product should have the calibration set saved from the sample
selection program.
2. Select the proper equation by double-clicking on it. The equation icon will turn blue.
Configuring the Method
1. Click on Regression/Edit Regression Method.
2. From the Edit Regression Method dialog box select Multiple Linear Regression, proper math
pretreatments, and the wavelength region.
Optionally enter the value for Wavelength Shift parameter (in nanometers). If the parameter is
different from 0, Vision will use its value to shift every spectrum in the calibration set left and
right along the wavelength axis. The resulting set of spectra, three times larger than the
original set, is used for calibration. Constituent values of the shifted spectra are identical with
these of the original spectrum.