# Re: Fault 1009

From: Gustavo A. Gutierrez <ggr03_at_IASTATE.EDU>
Date: Wed, 7 Feb 2007 18:36:31 -0600

Thanks Arthur for your quickly answer. I used ASREML version 1 (2002).

In order to identify which of the US structures is the problem, I ran the two
following models separately:

do ~mu line parity ys leg(acm,-1) !r leg(acm,1).anim ide(anim) !f mv
0 0 2
leg(acm,1).anim 2
2 acm US !GP !+3
0.27
0.082 0.07
1644 0 AINV
ide(anim) 1
0 0 I 1

do ~mu line parity ys leg(acm,-1) !r anim ide(anim).leg(acm,1) !f mv
0 0 2
anim 1
0 0 AINV 1
ide(anim).leg(acm,1) 2
1644 0 IDEN
2 0 US !GP !+3
0.1
0.01 0.01

Both models converged without any problem, and I got estimates for the variance
components, which I used as starting values to run a model with both random
regression components using the following code:

do ~mu line parity ys leg(acm,-1) !r leg(acm,1).anim ide(anim).leg(acm,1) !f mv
0 0 2
leg(acm,1).anim 2
2 acm US !GP !+3
1614
502 416
1644 0 AINV
ide(anim).leg(acm,1) 2
1644 0 IDEN
2 0 US !GP !+3
927
826 1204

The partial *asr file of the last model is:
Warning: Variance structures were modified in 8 instances to make them
positive definite. ASReml may have fixed the structure [flagged by
B] and may not have converged to a maximum likelihood solution.

Source Model terms Gamma Component Comp/SE % C
Variance 3830 1784 1.00000 5815.70 29.87 0 P
leg(acm,1).anim UnStruct 1 0.161400 938.653 29.87 0 B
leg(acm,1).anim UnStruct 1 0.811210E-01 471.776 29.87 0 B
leg(acm,1).anim UnStruct 2 0.416000E-01 241.933 29.87 0 B
ide(anim).leg(acm,1)UnStruct 1 0.533402 3102.11 29.87 0 B
ide(anim).leg(acm,1)UnStruct 1 0.250886 1459.08 29.87 0 B
ide(anim).leg(acm,1)UnStruct 2 0.120400 700.210 29.87 0 B

Covariance/Variance/Correlation Matrix UnStructured
0.1614 0.9900
0.8112E-01 0.4160E-01
Covariance/Variance/Correlation Matrix UnStructured
0.5334 0.9900
0.2509 0.1204

Gustavo

> Dear Gustavo,
>
> You wrote:
>
>
> I want to fit a random regression model for days open in dairy cattle,
> each cow
> had information from its first to fith calving (5 records per cow). The
> covariate is age at calving in months (acm). The code that I wrote is:
>
> anim * !P
> line * !A
> parity * !I
> ys * !I
> acm
> do
> newped.txt !MAKE #pedigree file
> datdofp1.csv !CSV #data file
> do ~mu line parity ys leg(acm,-1) !r leg(acm,1).anim ide(anim).leg(acm,1)
> !f mv
> 0 0 2
> leg(acm,1).anim 2
> 2 acm US !GP !+3
> .27
> 0.08 0.07
> 1644 0 AINV
> ide(anim).leg(acm,1) 2
> 1644 0 IDEN
> 2 0 US !GP !+3
> 0.16
> 0.147 0.21
>
> There is 1644 individuals in the pedigree file, and 766 with records (data
> file). Some cows have missing values for the response variable and some
> fixed
> effects. At this point I worked under the assumption of homogenous
> residual
> variance.
>
> The code was not converged, and I got the following message from *.asr
> file.
>
> NOTICE: 3 (more) singularities,
> 1 LogL=-8844.07 S2= 5572.3 1784 df
> 2 LogL=-8843.12 S2= 5583.3 1784 df : 2 components
> constrained
> 3 LogL=-8842.98 S2= 5571.3 1784 df : 1 components
> constrained
> Warning: EM updates for 1 positive definite US structure(s).
> 4 LogL=-8843.05 S2= 5520.3 1784 df : 2 components
> constrained
> 5 LogL=-8843.05 S2= 5520.3 1784 df
>
> Warning: Since fault 1009 occured during the last iteration
> results reported may be erroneous
> ????
>
> 13 1009 3830 3830 1000
> Finished: 06 Feb 2007 13:55:52.437 Unable invert R or G [US?] matrix
>
> What is the meaning of fault 1009?, and How I can solve it?
>
> ==================
>
> The problem is that one of the US structures has become singular and so
> ASReml can't proceed.
> You have not inducated which version of ASReml you are using. ASReml 2 is
> a little more
> robust in this regard, but the underlying problem is that you do not have
> much data, so even if
> your model is correct, sampling variation means it is not well estimated.
>
> So, what to do.
> 1) try and identify which of the US structures is the problem.
> 2) replace the US structure with
> CORUH .99 !GP
> 0.27 0.07 # Assuming it is the Genetic component
>
> This will fix the correlation at 0.99 which stops it being singular but
> keeps it positive definite.
>
>
>
> May you know Jesus Christ and His blessing in 2007,
>
> Arthur Gilmour, His servant .
>
> PS My Mixed Model Regression Mapping paper is now available at
> http://dx.doi.org/10.1016/j.csda.2006.12.031
>
> Personal website: http://www.cargovale.com.au/
>
> mailto:Arthur.Gilmour_at_dpi.nsw.gov.au, arthur_at_cargovale.com.au
> Principal Research Scientist (Biometrics)
> NSW Department of Primary Industries
> Orange Agricultural Institute, Forest Rd, ORANGE, 2800, AUSTRALIA
>
> fax: 02 6391 3899; 02 6391 3922 Australia +61
> telephone work: 02 6391 3815; home: 02 6364 3288; mobile: 0438 251 426
>
> ASREML 2 is now available from http://www.VSNi.co.uk/products/asreml
> The ASReml discussion group is at ASREML-L_at_dpi.nsw.gov.au
> To join it, mailto:arthur.gilmour_at_dpi.nsw.gov.au
> Cookbook: http://uncronopio.org/ASReml
>
> Proposed travel:
> ASReml workshop (Forestry) 16-18 April Tasmania
> Biom Brch meet 2-4 May
> ASReml workshop (Animal Breeding) 8-26 May South Africa
> AAABG 24-26 September Armidale
> <><><><><><><><><><><><><><><><><><><><><><><><><>
>
> This message is intended for the addressee named and may contain
> confidential information. If you are not the intended recipient or
> received it in error, please delete the message and notify sender. Views
> expressed are those of the individual sender and are not necessarily the
> views of their organisation.

Gustavo A. Gutierrez