Partial Least Squaresoverview how to example data see also
Partial Least SquaresStat > Regression > Partial Least Squares
Use partial least squares (PLS) to perform biased, non-least squares regression with one or more responses. PLS is particularly useful when your predictors are highly collinear or you have more predictors than observations and ordinary least squares regression either fails or produces coefficients with high standard errors. PLS reduces the number of predictors to a set of uncorrelated components and performs least squares regression on these components. You can use partial least squares when you have continuous or categorical predictors or a polynomial model.
PLS fits multiple response variables in a single model. Because PLS models the responses in a multivariate way, the results may differ significantly from those calculated for the responses individually. Model multiple responses together only if they are correlated. For more information, see [17] and [23].
Responses: Enter one or more columns containing the responses (Y).
Model: Enter the terms in the model. See Specifying a model - Partial Least Squares for more information.
Categorical predictors (optional): Enter the categorical terms that were already specified in the model.
Maximum number of components: Type the number of components to calculate or cross-validate. By default, Minitab calculates or cross-validates 10 components or the number of predictors, whichever is less. You should not enter more components than there are predictors.