48.1 Computing a PLS Regression (.NET, C#, CSharp, VB, Visual Basic, F#)
NMath Stats provides two classes for performing partial least squares (PLS) regression, PLS1 and PLS2:
● PLS1 is used when the responses, Y, in the model Y=XB+E consist of a single variable. In this case Y is a vector containing the n response values.
● PLS2 is used when the responses are multivariate. In this case Y is a matrix composed of n rows with each row containing the m response variable values.
Computing a PLS regression is accomplished by simply constructing a PLS1 or PLS2 instance. The basic parameters are:
● the matrix of predictor variables values
● the response variable values (a vector for PLS1 and a matrix for PLS2)
● an integer specifying the number of factors or components
For example:
Code Example – C# partial least squares (PLS)
DoubleMatrix A = ...
DoubleVector y = = ...
int numComponents = 3;
var pls = new PLS1( A, y, numComponents );
Code Example – VB partial least squares (PLS)
Dim A As DoubleMatrix = ...
Dim Y As DoubleVector = ...
Dim NumComponents As Integer = 3
Dim PLS As New PLS1(A, Y, NumComponents)
You can also invoke the Calculate() function on PLS1 or PLS2 to calculate a regression on an existing instance:
Code Example – C# partial least squares (PLS)
pls.Calculate( A, y, numComponents );
Code Example – VB partial least squares (PLS)
PLS.Calculate(A, Y, NumComponents)