[vsni.co.uk] Contact us
Author Message
Post new topic

  ASReml  ~  Modelling question

Posted: Wed Nov 19, 2008 5:48 pm Reply with quote
Joined: 18 Aug 2008 Posts: 25 Location: U.S.
I have several dozen field experiments for which I'm fitting the model:

yield ~ genotype (fixed) + y (random) + ar1(x):ar1(y)

where x,y indicate field coordinates.

When I make a scatterplot ( across all experiments) of the residual variance versus the variance component for y, I notice a striking pattern--either the residual variance is large and the variance component of y is small, or the opposite happens.does this suggest the model has too many terms and is trying to over fit the data?

View user's profile Send private message
Posted: Thu Nov 20, 2008 10:13 pm Reply with quote
Hi Kevin
It is hard to really provide an answer to this question without more information. Some of the questions that spring to mind are
1. You are fitting these as separate analyses, rather than doing a MET
2. What is the basic trial design, eg how many geno's how many replicates, what is the layout in terms of rows and columns (x and y in your syntax). I presume the trials are contiguous and equally spaced plots
3. are you fitting the random column term as a default and are there enough columns to warrant fitting an ar() model to the columns

If you would like me to look at the data then I would be happy to help and you can send a sanitised data frame to protect IP etc


Display posts from previous:  

All times are GMT
Page 1 of 1
Post new topic

Jump to:  

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum