cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Try the Materials Informatics Toolkit, which is designed to easily handle SMILES data. This and other helpful add-ins are available in the JMP® Marketplace
Choose Language Hide Translation Bar
imazarrasa
imazarrasa
Level I

How to set glm for gamma distributed dependent variable

Jan 30, 2020 05:06 AM (2113 views)

Hi,

My data is continuous and follow a ganma distribution. Which distribution and link function should I choose in the GLM platform? There is not option for ganma distribution..

I am using JMP 15.0.0

Thanks!

I.

0 Kudos
3 REPLIES 3
Mark_Bailey
Staff Mark_Bailey
Staff

Re: How to set glm for gamma distributed dependent variable

Jan 30, 2020 06:13 AM (2101 views)  |  Posted in reply to message from imazarrasa 01-30-2020

Use the exponential distribution and start with the reciprocal function as the canonical link.

0 Kudos
imazarrasa
imazarrasa
Level I

Re: How to set glm for gamma distributed dependent variable

Jan 30, 2020 07:50 AM (2093 views)  |  Posted in reply to message from Mark_Bailey 01-30-2020

Thank you very much for the answer.

And, in the case the data are continuous and don't follow a normal distribution nor a ganma distribution (even if they are log transformed), is it possible to apply a glm? 

Thanks again for your attention,

 

Inés.

0 Kudos
Mark_Bailey
Staff Mark_Bailey
Staff

Re: How to set glm for gamma distributed dependent variable

Jan 30, 2020 08:56 AM (2087 views)  |  Posted in reply to message from imazarrasa 01-30-2020

That question is not simple. The generalized linear model is very flexible with many choices from the family of natural exponential distribution models and link functions. So the choice of the best model depends on many things. The model assumptions and performance should be verified independently of the training of the model.

 

The GLM is decades old and has enjoyed a lot of attention, so there are many excellent resources created by experts. It is worth investing the time to study some of them if your modeling is important. For example, the on-line SAS documentation for SAS/STAT PROC GLM is a good place to start.

0 Kudos