hmsc-r / HMSC

GNU General Public License v3.0
99 stars 36 forks source link

How to use XSelect for variable selection #42

Open marissalee opened 4 years ago

marissalee commented 4 years ago

Dear HMSC developers -- First, thank you. I am excited about using your package for my research. Before I start using the many tools available, I would like to be able to use some sort of model selection procedure to help identify which environmental variables are important to retain in the XData matrix.

Based on the language in Hmsc() help, it sounds like the XSelect argument may help, but I do not fully understand what this does. For instance, in the examples you provide what do the q values specify? Is that a penalty for additional parameters?

Any additional references/help would be greatly appreciated. Thanks!

ovaskain commented 4 years ago

Dear Marissa,

Thanks very much for you interest on this.

The current R-package Hmsc includes some elements of model selection that are not yet described in any papers; they will be described in this book that should be out very soon: https://www.cambridge.org/fi/academic/subjects/life-sciences/ecology-and-conservation/joint-species-distribution-modelling-applications-r?format=PB

The features that relate to model selection include variable selection by slab-and-spike prior, reduced rank regression, and WAIC.

Unless you are in great hurry, maybe easiest to wait until the book is out; it includes the theory and R-scripts that demonstrate the use of these features.

If you are in great hurry in starting to use specifically these features, email me and I will try to help you get started.

Best,

Otso

On 25.3.2020 16.03, Marissa Lee wrote:

Dear HMSC developers -- First, thank you. I am excited about using your package for my research. Before I start using the many tools available, I would like to be able to use some sort of model selection procedure to help identify which environmental variables are important to retain in the XData matrix.

Based on the language in Hmsc() help, it sounds like the XSelect argument may help, but I do not fully understand what this does. For instance, in the examples you provide what do the q values specify? Is that a penalty for additional parameters?

Any additional references/help would be greatly appreciated. Thanks!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hmsc-r/HMSC/issues/42, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEIYMZUFEA73DGQZRJHRBDDRJIFKBANCNFSM4LTPDZ6A.