Re: ASReml-R and using weights with the ped() function

Hi,

I tried fitting a weighted analysis some months which worked fine
until I tried to use it in combination with a user defined inverse
(giv). From memory, I think this crashed R, and I presumed it was
some bug in the ASReml-R code. I haven't tried it with stand alone
ASReml.

Cheers,

Jarrod

On 10 Dec 2008, at 14:14, Bruce Southey wrote:

> kiwi wrote:
>> I've not received any comments about this one yet. There's a
>> further development:
>>
>> I mentioned the 'singularity' error I got when I ran this:
>> Code:
>> > discard.asr <- asreml(fixed = Discard_Total ~ 1 + Season,
>> random = ~ped(Vine, var = T) + Block,
>> data = model.df, na.method.X = "include",
>> ginverse = list(Vine = G.inv$ginv))
>> However, it didn't prevent convergence.
>> I mentioned that adding a weights argument causes an error to do
>> with a
>> Code:
>> Error : Abnormal termination
>> Singularity in Average Information Matrix
>> result.
>>
>> At that time, I considered it possible that the singularity had a
>> cabability of creating a problem with weighting, but otherwise it
>> was benign.
>>
>> I've found an instance where even without a singularity problem,
>> it's still a problem. The first one works fine; the second falls
>> over:
>> Code:
>> resp.asreml <- try(asreml(fixed = form,
>> random = ~ped(Vine, var = T) + Block,
>> data = model.df, na.method.X = "include",
>> ginverse = list(Vine = G.inv$ginv)))
>>
>> model.df$Wt <- rep(1, nrow(model.df))
>>
>> respW.asreml <- try(asreml(fixed = form,
>> random = ~ped(Vine, var = T) + Block,
>> weights = Wt,
>> data = model.df, na.method.X = "include",
>> ginverse = list(Vine = G.inv$ginv)))
>> The 'singularity' problem returns even though it would appear to
>> the ignorant that the same model was being tried.
>>
>> That leads me to conclude that it can't just be me.
>> So my question is this:
>> Has anyone ever got weights to work when using the ped() function?
>>
>>
>>
>>
>> -------------------- m2f --------------------
>>
>> Sent using Mail2Forum (http://www.mail2forum.com).
>>
>> Read this topic online here:
>> http://www.vsni.co.uk/forum/viewtopic.php?p=329#329
>>
>> -------------------- 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.
>>
> While you provide no data or output, you probably are lacking a
> suitable pedigree structure for your model and that there are terms
> confounded. Please check your data and pedigree file and then check
> your model as some terms can not be fitted as indicated. In
> particular:
> 1) Check that the model runs without including the pedigree.
> 2) If it does, run the model with only pedigree and the intercept.
> 3) If that model works slowing increase the model complexity with
> additional fixed and random effects
>
> Bruce
>
> 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.
>
>

-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.