Genetics User asking: Validation of Ridge Regression for estimated BLUP parameters?

Florian_Vogt

Hi all,

I am unfamiliar with the practices in  this field so am unsure about this question. A customer/prospect of JMP Pro 18 for Genetic capabilities asked this:

Can we do the validation of Ridge Regression for estimated BLUP parameters just to check the model sensitivity?

Appreciate any help here! @Chris_Kirchberg @Valerie_Nedbal 

Best

Florian

 

russ_wolfinger

@Florian_Vogt @Chris_Kirchberg  @chris_gotwalt1    @eclaassen   @Valerie_Nedbal @mia_stephens @SamGardner    I read this as a request for a Validation column to be added to Fit Model with Mixed Model personality.     

Two alternative ideas for now:

1. Create a validation column then a copy of Y in which all of the validation rows are set to missing.   Fit the RR-BLUP model using this new Y, save conditional predictions, then assess out-of-fold (single holdout) performance using the original Y.

2. The Torch add-in has a basic wide MixedModel available as an option under the Tabular Models pull-down menu that effectively does RR-BLUP.   It supports full repeated k-fold cross-validation and is fast.