topic Re: Principal Component Analysis in SAS Procedures

Principal Component Analysis

Davisonm1 — Wed, 19 Jun 2019 09:22:12 GMT

I want to perform a PCA on my data. I have one dependent variable and 31 independent variables. How do I select the variables after knowing the number of the Principal Components

Re: Principal Component Analysis

PaigeMiller — Wed, 19 Jun 2019 11:59:06 GMT

There is no such thing as a dependent variable in principal components analysis. So please explain further what you are trying to do here.

Re: Principal Component Analysis

Ksharp — Wed, 19 Jun 2019 12:22:24 GMT

Check PROC PLS , especially its example in documentation.

Re: Principal Component Analysis

PaigeMiller — Wed, 19 Jun 2019 13:08:44 GMT

@Ksharp wrote:

Check PROC PLS , especially its example in documentation.

Good point. If you want to do something like Principal Components on the X-variables to predict a Y-variable(s), then PLS is the way to go.

Re: Principal Component Analysis

Davisonm1 — Wed, 19 Jun 2019 13:08:59 GMT

I have 32 variables (1 dependent and 31 independent variables) and want to
perform data reduction (using PCA if this is the correct technique) to
remain with a few variables to use in explaining the variation in dependent
variable. I want to perform multiple linear regression and multiple
nonlinear regression analysis and determine which of the two will best
explain the changes in the dependent variable

Re: Principal Component Analysis

PaigeMiller — Wed, 19 Jun 2019 13:31:32 GMT

Definitely, this is a situation where you should use PLS and not PCA.

In PLS, you can determine the important variables in a number of ways, usually by looking at the loadings.