Analysis of a Split Plot & Random effect DOE with Generalized regression

[#1]

Is it possible to perform a good analysis of split plot/random effect DOE's with generalized regression? It can't handle random effects as is done with REML so is there a way to analyse correctly?

[#2]

I'm not sure exactly what question you are asking, but I suggest you read the following papers for ideas on how split-plots should be analyzed.

Box, G.E.P., Stephen Jones (1992), “Split-plot designs for robust product experimentation”, Journal of Applied Statistics, Vol. 19, No. 1

Bisgaard, Søren, (2000), “The Design and Analysis of 2 ^k-p X 2 ^q-r Split Plot Experiments”, Journal of Quality Technology, Vol. 32, No. 1, January

[#3]

Thanks! The Bisgaard article looks interesting. My question simply is if split plot result analysis is possible with generalized regression, a great tool but I don't know if applicable on DOE results with random effets. 

[#4]

A split-plot experiment designed by JMP Custom Design should save a Model table script that will launch an analysis with a valid mixed effects model.

[#5]

Can I use generalized regression with lasso or elastic net to analyse the split plot DOE results?

[#6 · Accepted solution]

Hi Frank,

Have you tried running the analysis in Generalized Regression? You will find that you get an error if you try to run a model with a random effect using Generalized Regression personality:

The simple answer is no.

I have had a few people ask me about model selection for designed experiment analysis when you have random effects (usually due to blocking, but the same applies for whole plot random effects). I recorded a short video tutorial that might be useful.

I hope this helps,

Phil

[#7]

I know random effects do't work in generalized but I thought there might be away to overcome so I guess not. 

[#8]

By random effects do you mean estimates of "pure error"? Random effects associated with which plot?  WP or SP? or both?  You should be looking at each plot separately.

[#9]

Hi @frankderuyck ,

No, there are "workarounds", as I mention in the video that I linked to. But they are only workarounds. There is no way to include random effects in the Gen Reg platform.

[#10]

Maybe it is better to use prior to REML Generalized regression with fixed random effects instead of Stepwise.