Re: rescaling design matrix
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: rescaling design matrix



re Peter Kube's question,
ASREML no longer runs in MSDOS real mode [since Dec 98 I think].
The Windows version can access all the memory that
is available.

re the following:


The problem is most likely related to the fact that the variance ratio
has gone to 50,000  i.e. the genetic variation
is 50 times larger than the residual variation.  

There is probably something odd in the data [or maybe you started
with this high value and it promptly failed.

With this data, you could just fit a FEMALE model

  oil ~ mu rep prov !r fem
  
  
  The variance ration (g) will then typically be less than 0.33
    So if half sib h2 is    4 g/(1+g) will be less than 1.
  
  However, in this case it may be more than 0.33
  
  In a tree model, the genetic variance is 4gS  and the residual is
  S(1-3g)   where S is the residaul variacne from the female model
  
  Now if g > 0.33, the residual   S(1-3g) becomes negative.
  
  ASREML will want to constrain the residual to be +ve so it has gone
  to a very small value and then been fixed there.
  
  So, in this case it is probably a problem with the model rather than the scale
  of the data.  If it was a scaling problem, you would rescale the oil
  variable [by say  !*0.01 transformation].  However, the scale of oil
  in this case looks fine.
  
  Arthur
  
  In wh
  
  When you fit the tree model, you expect a variance ration 4 times
  larger

 
> X-Authentication-Warning: lamb.chiswick.anprod.csiro.au: petidomo set sender 
to asreml-owner@lamb.chiswick.anprod.csiro.au using -f
> From: richardm@calm.wa.gov.au
> Date: Thu, 9 Mar 2000 15:54:40 +0800
> To: <asreml@chiswick.anprod.csiro.au>
> Subject: rescaling design matrix
> MIME-Version: 1.0
> 
> Hi all,
> 
> I am attempting to run a simple  univariate "tree" model. 
> 
>  Asreml has put out a warning message : WARNING:  1 singularities in AI 
matrix.
>   The suggestion is then made  that the design matrix be rescaled.  I am a
> little uncertain as to how this may be achieved.  Can this be done from within
> asreml?  Any pointers in the right direction would be greatly appreciated.
> 
> .as file is as follows
> 
> Merridin  P95 tree analysis
>  tree !p
>  fem 25 !I
>  male
>  rep 20
>  prov 3 !a
>  oil
> 
> 
> ellmer95.prn
> ellmer95.prn
> 
> oil~ mu rep prov !r tree
> 
> 
> relevent part of .asr  file is below:
> 
> 
>   ASREML [ 1 Jul 1999]  Merredin  P95 tree analysis                            
>          
>  09 Mar 2000 11:49:27.360   8.00 Mbyte    C:\asreml\ellmer95.as
>  A-inverse retrieved from ainverse.bin
>  PEDIGREE [ellmer95.prn ] has     1303 identities,    2581 Non zero elements
>   Reading ellmer95.prn  FREE FORMAT skipping  0  lines
>  Univariate analysis of oil                 
>  Using     1278 records [of    1278 read from    1278 lines of ellmer95.prn    
>     ]
>   Model term    Size Type    COL   Minimum    Mean      Maximum   #zero #miss
> 
>    1 tree       1303 Direct    1   2.000     663.9      1303.         0     0
> 
>    2 fem          25 Factor    2      1    13.2574         25         0     0
>    3 male          1 Covariat  3  0.000      0.000      0.000      1278     0
>    4 rep          20 Factor    4      1    10.6729         20         0     0
> 
>    5 prov          3 Factor    5      1     2.2926          3         0     0
> 
>    6 oil           1 Variate   6 0.1000      2.653      5.900         0     0
> 
>   Forming  1328  equations:  25  dense
>  Initial updates will be shrunk by factor    0.548
>  NOTICE:     2 (more) singularities,
>  LogL=-506.043     S2= 0.19031E-04   1256 df   0.5090E+05  1.000    
>  LogL=-506.043     S2= 0.19019E-04   1256 df   0.5093E+05  1.000    
>  Final parameter values                        50933.     1.0000    
>  WARNING:  1 singularities in AI matrix.
> 
>  Source                Model  terms     Gamma     Component    Comp/SE   % C
>  tree                   1303   1303   50933.5      0.968694      25.06   0 P  
>  Variance               1278   1256   1.00000      0.190188E-04   0.00   0 S  
>  WARNING: Code B - fixed at a boundary (!GP)
>                C - Constrained by user (!CON)
>                S - Singular Information matrix
>  S means there is no information in the data for this parameter.
>  Very small components with Comp/SE ratios of zero sometimes indicate poor
>   scaling.  Consider rescaling the design matrix in such cases
> 
>  Analysis of Variance       DF     F-incr      F-adj  StndErrDiff
>    4 rep                    20      35.83       8.03  0.1515    
>    5 prov                    2       6.99       6.99  0.2748    
> 
>                      Solution       Standard Error    T-value     T-prev
>    5 prov                
>  NB                     0.721356E-01   0.299888          0.24
>  WE                    -0.709678       0.295945         -2.40     -3.53
>    4 rep                 
>     
> 
> 
> thanks in advance
> 
> Richard
> =======================================
> Richard Mazanec
> Dept of Conservation and Land Management
> CALMScience Division
> Brain street, Manjimup, 6258.
> Phone +61 8 9771 7995    Fax +61 8 9777 1183
> Email: richardm@calm.wa.gov.au
> --
> Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml


<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD                  mailto:Arthur.Gilmour@agric.nsw.gov.au
Principal Research Scientist (Biometrics)            fax: <61> 2 6391 3899
NSW Agriculture                                           <61> 2 6391 3922
Orange Agricultural Institute             telephone work: <61> 2 6391 3815
Forest Rd, ORANGE, 2800, AUSTRALIA                  home: <61> 2 6362 0046

ASREML is still free by anonymous ftp from pub/aar on ftp.res.bbsrc.ac.uk
    or point your web browser at ftp://ftp.res.bbsrc.ac.uk/pub/aar/ 

To join the asreml discussion list, send the message  
     subscribe
mailto:asreml-request@chiswick.anprod.CSIRO.au

To send messages to the list, mailto:asreml@chiswick.anprod.CSIRO.au

Asreml list archive: http://www.chiswick.anprod.csiro.au/lists/asreml

                        <> <> <> <> <> <> <>
"Christ Jesus came into the world to save sinners"   I Timothy 1:15.
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>






--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml