nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Penalized Quantile and Expectile Regression

Muralidhara

Dear Team,

Two of our Academic Customers have requested for Penalized Quantile and Expectile Regression

Customer 1: Professor Dr Venkatesh, Alliance University

Customer 2: Professor Dr Abhijit, Myra School of Business

Currently these are not available with JMP17.

Thanks

Murali

Status: Needs Info

Comments

Sarah-Sylvestre

mia_stephens

Have split this into two Jiras.

We've had one other request for penalized quantile regression. According to the developer, we'd need to implement a new method in order to do this:

"We currently do quantile regression in genreg using iteratively reweighted least squares. It’s possible we could use IRLS and a data augmentation trick to do the lasso for quantile regression, but I think the interior point method is usually the go-to method for doing penalized quantile regression (especially if we wanted an elastic net penalty). So we’d probably need to implement an interior point solver if we decided to pursue this."

Exploring the possibility of implementing in JMP 19.

You can currently do expectile regression in the nonlinear platform. We've not seen any other requests to add, and it doesn't have as much momentum as quantile regression. Here's a paper, provided by the developer, that suggests that "expectile regression is not as sound as quantile regression".

mia_stephens

@Muralidhara can you please provide some additional information? For penalized regression, is LASSO sufficient, or is the customer also requesting Elastic Net and Ridge? A LASSO implementation might be relatively straightforward, but Elastic Net and Ridge will be more difficult, as they'd likely require different optimization techniques not currently available in JMP.