topic Re: What does zero factors mean in PLS cross validation? in Discussions

What does zero factors mean in PLS cross validation?

Zabarnss — Thu, 08 Jun 2023 16:28:28 GMT

PLS newbie here using JMP Pro 16. I have been using Kfold cross validation to select the number of factors in a PLS analysis. The table that results (Root Mean Press vs number of factors) begins with number of factors of 0. What does using zero number of factors mean? How can a PLS analysis be performed with zero factors?

Steve

Re: What does zero factors mean in PLS cross validation?

Zabarnss — Wed, 15 Mar 2023 13:57:54 GMT

Here is an example of the report:

Re: What does zero factors mean in PLS cross validation?

Mark_Bailey — Wed, 15 Mar 2023 17:28:39 GMT

The NIPALS method is iterative. It extracts one factor at a time. The first row represents the data itself. Subsequent rows represent the residuals after a factor has been extracted. See the statistical details for PLS.

Re: What does zero factors mean in PLS cross validation?

Zabarnss — Wed, 15 Mar 2023 17:38:58 GMT

Thanks for the reply Mark. So "zero factors" uses the original input data columns (X values) as "factors" (all of them)? If so, is this the equivalent to doing a multiple linear regression?

Re: What does zero factors mean in PLS cross validation?

Mark_Bailey — Wed, 15 Mar 2023 17:46:35 GMT

Did the JMP Help page I cite explain the computation?

Re: What does zero factors mean in PLS cross validation?

Zabarnss — Wed, 15 Mar 2023 17:51:00 GMT

The help was beyond my understanding of this topic. Does the first line show the PRESS before the first factor is extracted? If so, isn't this really a PLS with one factor (rather than zero)?

Re: What does zero factors mean in PLS cross validation?

Mark_Bailey — Wed, 15 Mar 2023 18:29:04 GMT

Yes, the first PRESS is before any factor is extracted. It is not for a PLS model with one factor. The example you show above suggests selecting just one factor because the p-value for T square is above 0.1 for the first time when 1 factor is extracted.