Include random effects in forward stepwise parameter selection

[frankderuyck]

Many DOE's need to be peformed over several days so considering and separation of random time variation in DOE analysis is an issue. As stated by Goos & Jones Randomized blocks in DOE are very useful to absorb random variation so fixed effects can be precisely estimated.

However in the JMP forward stepwise parameter selection module random effects can't be included; all blocks, also random time effect blocks, need to be treated as fixed effects. In forward selection it would be a major imporvement to also allow random effects so the old cumbersome backward selection can be avoided.

[JesperJohansen]

Also allowing random effects in the Fit All Possible Models feature would be very desirable.

[Ryan_Gilmore]

We are archiving this request. If this is still important please comment with additional details and we will reopen. Thank you!

[frankderuyck]

Yes still important to know how to handle random effects (random block, whole plot..) in the stepwise regression and also in generalized regression. How is a split plot DOE now analysed with generalized regression?

[JesperJohansen]

I still think it is important.

I dont know what additional details are needed?

Currently, in Fit Model, the Stepwise/Fit All Possible Models features do not allow models with random effects. It would be great if it did.

[frankderuyck]

Generalized regression is a great tool and very useful for analysing DOE! Maybe it is difficult to handle random effects/split plots with shrinkage methods, if so is there a way to handle this with generalized regression?

[Sarah-Sylvestre]

@frankderuyck Thank you for your response! I have changed the status back to 'Acknowledged'.  We have captured your request and will take it under consideration.  No further details are needed at this time. 

[Sarah-Sylvestre]

@frankderuyck - We have reviewed your request and think it will be a great addition to JMP. This request is in the queue of items to work on in a future release.

[datadad]

Agree. Design expert can do this. JMP is behind

[hogi]

candidates of the Wish List Prioritization Survey - pilot run:
Introducing the JMP Wish List Prioritization Survey!