topic Re: Strip-Split-Plot in Statistical Procedures

Strip-Split-Plot

GBR2015 — Thu, 25 Jun 2015 02:06:32 GMT

Hello,

I have a Strip-Split-Plot design with SAMPLING on the lowest plot. I already have done hard work researching and did not get final conclusion if I even must use a subsample error term and if what is correct error for the sub-sampling in these design.

I wrote the code below, and would be great full if someone could help me suggestim a better one.

PlotA=main plot, and within each PlotA I have a strip-plot with Row factor and Col factor. (With SUBSAMPLING)

PROC GLIMMIX data=dataset plots=studentpanel;

Class Block PlotA Row Col;

Model Y=PlotA | Row | Col / ddfm=satterthwait;

Random Block(PlotA) Row*Block(PlotA) Col*Block(PlotA) Row*Col*Block(PlotA);

run;

Thank you.

Re: Strip-Split-Plot

SteveDenham — Fri, 26 Jun 2015 17:13:45 GMT

See this thread for a strip-split plot model

https://communities.sas.com/thread/79513

Steve Denham

Re: Strip-Split-Plot

GBR2015 — Fri, 26 Jun 2015 21:46:49 GMT

Thanks Steve Denham

I checked the link and I believe that my code is correct.
Since I have a strip-plot within a simple randomize complete block (RCB) with Factor 'A' with ''a' levels, the interraction 'Block*A' become the "blocks" for the strip-plot, is it OK?

Thank you

Re: Strip-Split-Plot

SteveDenham — Mon, 29 Jun 2015 12:57:40 GMT

That sounds correct, and also sounds a lot like an example in SAS for Mixed Models, 2nd ed., where an extensive strip plot example is worked out.

Steve Denham

Re: Strip-Split-Plot

sld — Mon, 29 Jun 2015 19:09:11 GMT

If by "randomized complete block (RCB) with Factor 'A' with 'a' levels", you mean that each level of A is applied to a whole plot (identified as "Block*A") within each block--in other words, you have 'a' whole plots in each block--and that the row strips and the column strips are entirely contained within each whole plot, then ideally you would want

random block

block*a

block*row

block*a*row

block*col

block*a*col

block*row*col

block*a*row*col;

Lots of variance estimates! So the full list of terms might not be feasible.

Your original RANDOM statement would be correct if levels of factor A were randomly assigned to blocks rather than whole plots within blocks.

Susan

Re: Strip-Split-Plot

SteveDenham — Tue, 30 Jun 2015 16:13:20 GMT

Susan is right--I read it as a split-split plot, not a strip-split plot. Mark her answer as correct.

Steve Denham