Re: Residuals from AR1xAR1 one model

If you fit the units term as well in the model, this is closer to the lack
of fit definition of the residuals. I'm sorry, I don't know the equivalent
in ASReml-R.

We find that this term is very important for forestry trials representing
random tree to tree variation (still there in clonal trials), and is also
fitted quite a bit in agricultural variety trials where it is usually said
to represent measurement error.

So fitting the AR1 and units in R will give you an R that reflects both
independent and spatially structured residuals.

You could also fit the AR1 in the G part of the model and leave the R purely
as independent (lack of fit type) error (covered in the ASReml manual page
113) if such a complex R worries you, but the result is the same.

  _____

From: ASReml users discussion group [mailto:ASREML-L_at_DPI.NSW.GOV.AU] On
Behalf Of kwstat
Sent: Friday, 26 September 2008 4:44 AM
To: ASREML-L_at_DPI.NSW.GOV.AU
Subject: Residuals from AR1xAR1 one model

I usually think of residuals as being the difference between the observed
values and the modeled/fitted values. Residual plots can then be used as
diagnostics for the fit of the model.

When I fit a model with an AR1xAR1 spatial structure in the residual
covariance matrix, the above philosophy is not quite right. For example, if
I look at a heat map/level plot of the residuals, I usually see a smoothish,
undulating trend. This is the structure imposed by the correlation of the
residuals, and is not structure in the data that that has not been
adequately modeled.

There is a resid() function in ASREML-R, but these aren't really deviations
from the model in the usual sense, are they?

What is the right way to think about these quantities?

Kevin Wright

-------------------- m2f --------------------

Sent using Mail2Forum (http://www.mail2forum.com).

Read this topic online here:
http://www.vsni.co.uk/forum/viewtopic.php?p=243#243

-------------------- m2f --------------------

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 Sun Sep 26 2008 - 06:53:55 EST