turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- factorial split-plot with a strip factor

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-17-2015 09:43 AM

Dear SAS Community,

My design is a factorial split-plot; main plots consisted of a 2-factor factorial design (nem and trt). Main plots were divided into 2-subplots, each cropped to a different cultivar. Since a different cereal was cropped every year in the plots, to compare the response var (pop) between cereals, cereal was analyzed as strip factor within the split plot model.

I would like to know is this model is correct:

proc mixed data=one;

class blk nem trt cultivar cereal;

model pop=nem|trt|cultivar|cereal/ddfm=kr;

random intercept nem*trt cereal nem*cereal trt*cereal nem*trt*cereal cultivar*cereal nem*cultivar*cereal trt*cultivar*cereal/subject=blk;

run;

I would greatly appreciate your comments!

Thank you,

Caroline

Accepted Solutions

Solution

06-18-2015
02:10 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to palolix

06-18-2015 02:10 PM

The random terms all look correct for a strip plot, I just worry about calculating four-way interactions with any degree of precision unless you have a LOT of data. IF the model converges, then the tests should have the right degrees of freedom and compare the correct quadratic forms. Seems a big IF to me, though.

Steve Denham

All Replies

Solution

06-18-2015
02:10 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to palolix

06-18-2015 02:10 PM

The random terms all look correct for a strip plot, I just worry about calculating four-way interactions with any degree of precision unless you have a LOT of data. IF the model converges, then the tests should have the right degrees of freedom and compare the correct quadratic forms. Seems a big IF to me, though.

Steve Denham

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to SteveDenham

06-18-2015 02:55 PM

Thank you very much for your help dear Steve!! it did work and also converged for all dependent variables. Since I do not have missing values maybe I should use satherwaite instead of kr....do you agree?

Thank you Steve!!

Caroline

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to palolix

06-18-2015 03:01 PM

Hi Caroline,

Too many variance components with different degrees of freedom. Be pragmatic about this, how much gain (or loss) in power do you get from the shrinkage part of KR, as opposed to the combining of variances using Satterthwaite's method? Just look at the standard error of the difference for trt or nem as a guidance.

Steve Denham

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to SteveDenham

06-18-2015 03:29 PM

the SE for the difference between nem or trt is the same value (0.06268) when using kr, satterwaithe or nothing.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to palolix

06-18-2015 03:32 PM

Not surprised with completely balanced data, and no repeated measures.

Steve Denham

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to SteveDenham

06-18-2015 03:44 PM

Appreciate your comments Steve. Thank you!

Caroline