Re: residual covariance (repeated measures)

Thanks Brian

This is along the lines I was thinking. As I understand, you indicate
that there is no way to set up a covariance structure in the R matrix if
the data observations are unbalanced without inserting records with
missing values.

Sorry didn't want the email to get bogged down in long description of
issues but will try to give a bit more detail next time.

Regards

Dr Craig Hardner

Research Fellow

School of Land, Crop and Food Sciences

University of Queensland

St Lucia 4072 Queensland, Australia

ph: +61 7 3346 9465

email: craig.hardner_at_uq.edu.au <mailto:craig.hardner_at_uq.edu.au>

From: ASReml users discussion group [mailto:ASREML-L_at_DPI.NSW.GOV.AU] On
Behalf Of brian.cullis_at_INDUSTRY.NSW.GOV.AU
Sent: Wednesday, 23 September 2009 10:20 AM
To: ASREML-L_at_DPI.NSW.GOV.AU
Subject: Re: residual covariance (repeated measures)

Dear craig
You wrote
  
I am trying to figure out the correct code in R to a repeated measure
analysis on unequal numbers of observations with 1 trait measured on 2
occasions. 417 plants were assessed in year 2 and 590 plants were
assessed in year 2. 81 plants were assessed in both years. We are trying
to build a var-cov matrix at both G and R, with a relationship matrix at
G (about 15 parents used in crossing scheme).
  
The data is listed in long format and sorted by Year, i.e.
Year 1 data
Year 2 data
  
I am wondering, is it possible to specify a multivariate rcov matrix
without having to put dummy records in for both traits to balance the
data set?
  
When I use
  
rcov=~diag(year):units
  
or
  
rcov=~diag(year):ide(Ind)
  
I get errors from ASreml expecting balanced data set
  

I reply
With the little information I have, ie no background etc which is always
scary to give advice, I would trick ASReml to fit G and R both as G
structures, and fix S2/dispersion to a very small number, eg .0001. This
is simple and quick and then you can specify
both ped(tree) and residual as G structures.

Eg two years, aka two traits has

random=~us(trait):ped(tree) + us(triat):tree,
family=asreml.gaussian(dispersion=.0001),....
where trait:tree and units are equivalent in the sense that each unit is
uniquely specified by the tree and the trait combo

HTH
 

warm regards

Brian Cullis|Research Leader, Biometrics &
Senior Principal Research Scientist |
Industry & Investment NSW | Wagga Wagga Agricultural Institute | Pine
Gully Road | Wagga Wagga NSW 2650 |
PMB | Wagga Wagga NSW 2650
T: 02 6938 1855 | M: 0439 448 591 | F: 02 6938 1809 | E:
brian.cullis_at_industry.nsw.gov,au
W: www.industry.nsw.gov.au |

Visiting Professorial Fellow
School of Mathematics and Applied Statistics
Faculty of Informatics
University of Wollongong

Professor,
Faculty of Agriculture, Food & Natural Resources
The University of Sydney

Adjunct Professor
School of Computing and Mathematics
Charles Sturt University

This message is intended for the addressee named and may contain
confidential information. If you are not the intended recipient, please
delete it and notify the sender. Views expressed in this message are
those of the individual sender, and are not necessarily the views of
their organisation.

 
Received on Thu Sep 23 2009 - 10:25:41 EST