Subscribe Bookmark



Jun 23, 2011

Roundtable Experience at JMP User Conference

It is hard to believe that the JMP User Conference is over for another year. So much happened in just one week! I left with lots of new ideas.

Some of these ideas occurred to me in the lunch time roundtable discussions. I was fortunate to moderate one about design of experiments (DOE) on both days of the conference. Our expert was Dr. Stuart Hunter. On the second day, we had a surprise guest join us: Dr. Doug Montgomery, the keynote speaker!

They were asked for their opinion about what the big new areas will be for DOE. Both experts agreed on three fronts: designs for computer experiments, non-linear models, and generalized linear models (GLM). They feel that theory is at hand to address these designs but to put it in to practice will take good software.

JMP offers non-linear design and space-filling designs and gaussian process models for computer experiments. JMP can also fit many GLM models. So it seems that the experts and JMP foresee the same future.

1 Comment
Community Member

Yves Chandon wrote:

I am glad to read those comments on DOE evolution. One of the big topics is to model the variability as well a as the mean. The Logvariance model has been a significant progess for that. Then the Generalized Linear Model has been introduced. In fact it would be great to be able to use it to model the variance with a gamma distribution, a Log link, and a scale parameter for the sample size as described in "Process Characteriztion and Optimization", Tobin et all, Addison-Wesley/SSRI. Are there any plan to include that in future JMP release?

Yves Chandon