[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


> From: Trevor Hancock <>
> Arthur(et al),
> I have two further matters.  The first is rather trivial but confirmation 
> of my observation would be appreciated.  I have some data which has 
> missing values for the response variate yield ( indicated by *'s ).  When 
> I read it into ASREML it indicates the #miss is zero.  More careful 
> inspection indicates the # of records retained = total observations - 
> #missing etc and it appears that the *ing values have been recognised 
> appropriately, as stated in the manual.  Is this in fact correct and there 
> is a minor bug in #miss shown??

Missing values are treated in various ways.  In your job, they are
not being fitted and so ASREML automatically deletes the records
containing missing values in the trait being analysed.  
The  #MISS column records the Number missing in the
records that are retained.  So, if you had another trait with a different
pattern of missing values, you could see how many were missing
in the records where the first trait was present.

If you include  the special  'mv'  model term [ it must be
fitted as a fixed effect] in the model, or if you are doing
a multivariate analysis, records with * in the dependent variable
are not dropped.

> My second ? is less trivial (I think). I have data where I expect a model 
> with common intercept but different slope would seem likely given the 
> biology.  That is, at time zero it would be expected all treatments start 
> at the same value and gradually separate depending on the treatments.
> Thus I fit 
> 	yield ~ mu variate #for a common line
> then	yield ~ mu variate factor.variate #common intercept different 
> slope.  This second model appears strange and in fact bears little 
> similarity to REML (in GENSTAT).  Whereas the model
> 	yield ~ mu variate factor factor.variate #different intercept and 
> slope,
> and the initial model appear reasonable and are consistent with REML.
> Is there any reason why this second model should "upset" ASREML ??
> Note eventually I want to fit a much more complicated model involving 
> pol(), spl() and other factors but if the simple 2nd model doesn't give 
> me what I expect I'm a little reluctant to head for the "deep water" and 
> beyond, without some advice from the lifeguard(s).
There is no reason 
   yield ~ mu variate factor.variate 
should upset ASREML
   There is a singulatity and GENSTAT might handle it differently
so that the fitted model is not the same.

In ASREML, the coefficient for 'variate' will actually be
the slope for factor_1
and the coefficients for factor.variate will be differences in slope from
the slope of factor_1.

You could fit the model as
   yield ~ mu factor.variate
to avoid the singularity 
> Cheers,					Trevor.
>  ************************************************************************
>  TREVOR  W  HANCOCK, Biometry, Dept. of Plant Science, Univ. of Adelaide.
>  Waite Institute, PMB 1 Glen Osmond, South Aust.,AUSTRALIA.     5064
>  Tel:   (08) 8303 7288    International:  61 8 8303 7288
>  Fax:   (08) 8303 7109    International:  61 8 8303 7109
>  email:
>  ************************************************************************
Arthur Gilmour PhD                           email:
Senior Research Scientist (Biometrics)                 fax: <61> 2 6391 3899
NSW Agriculture                             telephone work: <61> 2 6391 3815
Orange Agricultural Institute                         home: <61> 2 6362 0046
Forest Rd, ORANGE, 2800, AUSTRALIA         

ASREML is currently free by anonymous ftp from pub/aar on
Point your web browser at **NOTE CHANGE**
in the IACR-Rothamsted information system 

To join the asreml discussion list, send the message  

The address for messages to the list is

   Where is He who has been born King of the Jews? For we have seen His star 
   in the East and have come to worship Him.  Matthew 2:2