topic Re: 4 Factorial-Split Plot Proc mixed in Statistical Procedures

4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 12:57:41 GMT

Hi, does anybody knows how to code a 4 Factorial-Split Plot design with proc mixed?

Thank´s a lot!

Caroline

Re: 4 Factorial-Split Plot Proc mixed

PaigeMiller — Thu, 07 Feb 2013 13:05:47 GMT

There are examples of using PROC MIXED for split plots in both the online SAS help, plus in the book "SAS System for Mixed Models" by Littell, Milliken Stroup and Wolfinger.

You really haven't given us enough details about the design for us to give more explicit instructions.

Re: 4 Factorial-Split Plot Proc mixed

SteveDenham — Thu, 07 Feb 2013 13:18:07 GMT

I could give it a shot, but details of the design, including the nature of all variables, are absolutely a minimum requirement, as per .

Steve Denham

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 13:21:17 GMT

Thank you very much for your answer Paige!

I already cheked it on SAS help, but there are only examples with 2 factors, but not with 4. In my design, I want to test the effect of the following variables:

Block

Year (3)

Nematodes (2 species)-wholeplot

Treatment (2)-Plot

Cereal Cultivar (3)-Subplot

Here is how I do it with 3 factors, but I do not know how is the code when I want to add another factor (year)

Proc mixed data=one;

class blk nem trt cultivar;

model lgsprCysts100=

nem

trt

trt*cultivar

cultivar

nem*cultivar

trt*cultivar

nem*trt*cultivar/ddfm=Satterth;

random blk blk*nem blk*nem*trt;

Re: 4 Factorial-Split Plot Proc mixed

PaigeMiller — Thu, 07 Feb 2013 13:31:53 GMT

Is Year crossed with the other factors?

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 13:37:57 GMT

I guess, because each year a different cereal was planted. I want to compare the dependent variable between 3 years.

Re: 4 Factorial-Split Plot Proc mixed

PaigeMiller — Thu, 07 Feb 2013 13:46:46 GMT

If each year a different cereal was planted, then cereal is nested within year, but year could still be crossed with treatment and nematodes. Could you confirm?

These are the types of details that we require if we are going to provide a meaningful and correct answer.

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 13:56:49 GMT

Yes, I guess cereal could be nested in year, and year could also be crossed with treatment and nematodes.

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 16:49:09 GMT

I think maybe I should analize it as a Split-split plot model in order to include the year, which should be include before block and the rest of the factors. What do you think?

Re: 4 Factorial-Split Plot Proc mixed

PaigeMiller — Thu, 07 Feb 2013 17:06:32 GMT

The problem is that you have to analyze the design in an appropriate manner. So if Year is crossed with the other factors, then it can't be a split-split plot (I don't think).

So, I think this is the analysis

Year and Nematode and their interaction is the whole plot, treatment is plot, cereal cultivar nested within Year is subplot.

So this is my analysis

proc mixed;

class year blk nem trt cultivar;

model lgsprCysts100=nem|year trt trt*nem trt*year trt*nem*year cultivar(year) cultivar(year)*nem cultivar(year)*trt cultivar(year)*new*trt;

random blk blk*trt blk*trt*nem blk*trt*year blk*trt*year*nem;

run;

Maybe I'm way off. What do you think, ?

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 07 Feb 2013 17:21:28 GMT

Thanks a lot Paige! its look very good. Maybe I said it wrong, but block should be the whole plot, because each block includes the 3 nematodes, the 2 different treatments and the cultivars, so nematode and treatment should be the plot, and cultivar the subplot.

Re: 4 Factorial-Split Plot Proc mixed

SteveDenham — Mon, 11 Feb 2013 14:14:34 GMT

Sorry to be so slow in replying. That model ought to be OK. Now I'm going to worry more about the dependent variable, and whether this should be done in GLIMMIX with an untransformed DV and an appropriate distribution for counts of cysts.

If it were in GLIMMIX, with a negative binomial distribution for overdispersed counts, we could also get an appropriate covariance error structure to model any time dependence over years.

Steve Denham

Re: 4 Factorial-Split Plot Proc mixed

palolix — Thu, 14 Feb 2013 09:10:08 GMT

Thank you very much for your reply Steve! I think this model should be also ok.

Proc mixed data=one;

where nem>1;

class blk nem trt crop date;

model lgsprCysts100=

nem|trt|crop|date/ddfm=Satterth;

random blk blk*nem blk*nem*trt*crop;

Re: 4 Factorial-Split Plot Proc mixed

palolix — Fri, 15 Feb 2013 17:15:37 GMT

Hi Steve!

I have a Split plot design but with 4 factors; nem, trt, crop and date. Until now I have only done split-split designs, but what comes next when you have to add a forth factor. Is this code ok??

Thanks!!

blk

nem

blk*nem

trt

nem*trt

blk*nem*trt

crop

blk*crop

nem*crop

trt*crop

blk*nem*crop

nem*trt*crop

date

blk*date

nem*date

trt*date

crop*date

blk*nem*trt*date

nem*trt*crop*date;

random blk blk*nem blk*nem*trt blk*crop blk*nem*crop blk*date blk*nem*trt*date/test;

run;

Re: 4 Factorial-Split Plot Proc mixed

SteveDenham — Mon, 18 Feb 2013 12:35:15 GMT

Caroline,

I think there may be some missing factors. Let me see if the design is correct:

Fixed effects: nematode type, treatment, crop

Random effect: block

Repeated effect: date

I assume that there is a single measurement per date for the nem*trt*crop*block cell.

If this is the case, it looks like the following might be what you should consider:

proc mixed;

class nem trt crop blk date;

model nem|trt|crop|date;

random intercept nem|trt|crop/subject=blk;

repeated date/subject=nem*trt*crop*blk type=<insert appropriate type here depending on spacing of dates, cs or csh for uneven spacing, ar(1) or arh(1) for regular spacing;

run;

Appropriate options should be added for solutions, etc., as well as LSmean or LSestimate statements as needed.

Steve Denham

Re: 4 Factorial-Split Plot Proc mixed

palolix — Tue, 19 Feb 2013 13:56:07 GMT

Hi Steve,

thank you very much for your help!

ja, there is a single measurement per date, and and there is a regular spacing for date (harvest 2010, harvest 2011 and harvest 2012). The thing is that is not a repeated measures model, because the crop are not the same every year, so the conditions are not the same every year. How would you code this model with proc glm?

Thanks a lot!

Caroline

Re: 4 Factorial-Split Plot Proc mixed

SteveDenham — Wed, 20 Feb 2013 16:48:49 GMT

Hi Caroline,

I would never code it with proc glm. See Walter Stoup (2013) Generalized Linear Mixed Models.

PROC GLM just does not do a good job of analyzing repeated measures or split plot models.

The fact that the crop is not the same across all dates does present a problem in terms of estimability of means. There will be "missing" cells by design. To get around this, I would fit a cell means model, and use LSMESTIMATE statements to derive lower order means and comparisons.

proc glimmix;

class nem trt crop blk date;

model lgsprCysts100=nem*trt*crop*date/noint; /* This fits a cell means model */

random intercept nem|trt|crop/subject=blk;

random date/ residual subject=nem*trt*crop*blk type=arh(1); /* This may be problematical to fit, due to the small number of levels for date, and incomplete subjects. If so, try type=chol */

run;

The LSMESTIMATE statements will need to be added later. I also have a hunch that the dependent variable is a count variable, and rather than a log transform prior to analysis, a log link should be used so that the dependence of the variance on the mean can also be captured.

Steve Denham

Message was edited by: Steve Denham

Re: 4 Factorial-Split Plot Proc mixed

palolix — Wed, 20 Feb 2013 17:04:47 GMT

Thanks a lot Steve!

I used to run my models (also split plot models) as a first run with proc glm in order to pool, so I can get a clean model when using proc mixed. In this case, I think I do not have a repeated measures, because not only the crops are different every year, but also the planting dates, so the circunstances differ every year.

Caroline