Dear James,
I'm a bit slow! You are over the TASMAN.
There is no problem with your coding.
Explanation of parameter values printed out is:
When ASREML sees
a model ~ sex age !r animal dam ide(dam) !f cg
It initializes the first 6 positions in the vector of parameters
so ide(dam) is in cell 5 and cg in cell 6 [its value is zero since it
is a fixed effect]
Then ASREML alolocates a cell for the error variance [7] and then for
the US structure [animal dam] cells 8-10.
What is printed in the LogL lines is cells 5 - 10 of this vector
so that explains where the zero comes from.
In the printout of variance parameters, the redundant [undefined] positions
are ignored.
When it comes to the .pin file, the defoned terms are taken from the
> Source Model terms Gamma Component Compnt/StndErr
> ide(dam) 27066 7268 0.147784 0.147784 6.57
> animal US=UnStr 1 -0.766052E-02 -0.766052E-02 -2.69
> animal US=UnStr 1 -0.111642E-02 -0.111642E-02 -0.17
> animal US=UnStr 2 0.278432E-01 0.278432E-01 1.39
table so only 4 items are defined. [Conseqently the .vvp file shp`ould
have only 10 values.]
You are correct in interpreting the components
However, you now catch me out. Because the Variance of 1 is not in this table,
there is no way of including it in the .pin file calculations [ I do not
have a sysntax for addind a constant to a variance] other than
appending a 5th row
Residual 0 0 1. 1. 0.
to the table in the .asr file and appending 5 zeros to the .vvp file.
I'll address this deficiency.
Concerning the analysis you have done, It seems you probably have no
direct genetic effects on survival, only maternal ones.
Arthur
>
> Arthur,
>
> I feel a bit silly having to write to you again, I have almost got the asreml
sorted out but have
> one or two little questions. Please feel free to put the reply on the ASREML
list if you do not
> think the question is too silly and may help someone else. BTW you were right
before with the
> previous question I had - we had one errant animals missing from our ped file.
>
> I am using the following input file:
>
> computer> :more sup1.as
> ASREML analysis of surv at 48 hrs old
> animal !P
> sex 2 !A
> age 10 !A
> cg 18 !A
> dam !P # has genetic group codes in here too
so no missing dams
> LS48 1 !M-1.00 # coded as 0 = dead 1 = alive -1.00 =
missing
> ped.dat !GROUPS 43
> data.dat
> LS48 !bin !probit ~ sex age !r animal dam ide(dam) !f cg
> 0 0 1
> animal 2
> 2 0 US 0.3 0.0 0.3
> animal
> computer>:
>
> I am using probit because the Logit does not converge ?? but the probit does
> The model i am trying to fit is
> LS = sex + dam_age + contemp_group + additive genetic + maternal genetic +
dam
> permanent environment (random)
>
> With the output I get :
>
> computer> :more sup1.asr
> ASREML [16 Apr 1998] ASREML analysis of surv data
>
> Thu May 7 17:58:54 19981024.00 Mbyte sup1.asr
>
> A-inverse retrieved from ainverse.bin
> PEDIGREE [ped.dat ] has 27066 animals, 95949 Non zero elements
> Reading data.dat FREE FORMAT skipping 0 lines
> Univariate analysis of LS48
> Using 22703 records [of 22703 read from 22703 lines of data.dat
> ]
> Model term Size Type COL Minimum Mean Maximum #zero
#miss
> 1 animal 27066 Direct 1 3985. 0.1541E+05 0.2707E+05 0
0
> 2 sex 2 Factor 2 1 1.4969 2 0
0
> 3 age 10 Factor 3 1 2.0807 10 0
0
> 4 cg 18 Factor 4 1 8.4273 18 0
0
> 5 dam 27066 Direct 5 45.00 7609. 0.2326E+05 0
0
> 6 LS48 1 Variate 6 1.000 0.8884 1.000 2533
0
> 7 ide(dam) 27066 Factor 5 45 7608.8208 23264 0
0
> 2 US=UnStr 0.30 0.00 0.30
> 27066 Ainverse
> Structure of animal has 54132 levels defined
> Forming 81229 equations: 13 dense
> Initial updates will be shrunk by factor 0.010
> Data distribution and link: Binomial; Probit P=Normal Integral(XB)
> Deviance adjust= -19139.432356986621
> NOTICE: 19845 (more) singularities,
> LogL=-23519.1 S2= 1.000 22651 df 0.10000 0.00000 1.00000
> 0.30000 0.00000 0.30000
> Deviance adjust= -11356.4644239385
> LogL=-22110.3 S2= 1.000 22651 df 0.14888 0.00000 1.00000
> -0.00766 -0.00112 0.02789
> Deviance adjust= -13120.537311384705
> NOTICE: 19846 (more) singularities,
> LogL=-22071.9 S2= 1.000 22651 df 0.14806 0.00000 1.00000
> -0.00767 -0.00111 0.02784
> Deviance adjust= -13131.675942170163
> LogL=-22082.8 S2= 1.000 22651 df 0.14779 0.00000 1.00000
> -0.00766 -0.00111 0.02784
> Final parameter values 0.14778 0.00000 1.00000
> -0.00766 -0.00112 0.02784
> Variance heterogenity factor 0.58
>
> Source Model terms Gamma Component Compnt/StndErr
> ide(dam) 27066 7268 0.147784 0.147784 6.57
> animal US=UnStr 1 -0.766052E-02 -0.766052E-02 -2.69
> animal US=UnStr 1 -0.111642E-02 -0.111642E-02 -0.17
> animal US=UnStr 2 0.278432E-01 0.278432E-01 1.39
> Solution Standard Error T-value T-prev
> 2 sex 1 9.02 9.02 [ DF F_inc F_all ]
> M -0.706125E-01 0.235071E-01 -3.00
> 3 age 8 6.50 6.54 [ DF F_inc F_all ]
> 03 0.191854 0.354455E-01 5.41
> 04 0.628799E-01 0.511734E-01 1.23 -3.14
> 05 0.278776E-01 0.676825E-01 0.41 -0.55
> 06 -0.322999E-01 0.951852E-01 -0.34 -0.65
> 07 -0.184251 0.189072 -0.97 -0.79
> 08 0.363875 0.307400 1.18 1.64
> 10 0.117837E-01 5257.86 0.00
> 01 1.31682 0.379077 3.47 0.00
> 1 animal 27048 effects fitted
> 5 dam 27039 effects fitted
> 7 ide(dam) 7268 effects fitted
> 4 cg 18 effects fitted
> 1881 possible outliers: see .res file
> Finished: Thu May 7 19:32:46 1998 LogL Converged
> computer> :
>
>
> now I am confused as to the order that these effects are being fitted:
> ie
> Final parameter values 0.14778 0.00000 1.00000
> -0.00766 -0.00112 0.02784
>
> I assumed the 0.14778 is the dam pe effect (not augmented with NRM)
> I have no idea what the 0.0000 is ?????
> The 1.000000 is the error variance which is fixed to 1.00000 automatically
with the !bin
> ie !s2=1.00
> The -0.00766 is the add genetic
> The -0.00112 is covariance between add genetic and maternal genetic
> The 0.02784 is the mat genetic effects
>
> So what is the 0.00000 effect ?
> Have I mucked up the model specs to create this bogus random term?
> And what are the order numbers of the variance components so I can make up a
sensible .pin
> file to calculate correlations etc?
> Should the .pin be set up as per usual with a residual variance of 1.00 to
give standard
> heritability and rgs on the transfromed scale?
> And finally what do the Compnt/StndErr 's mean on the transformed scale?
>
> Many thanks and terribly sorry to bother you again.
>
> skerry
>
> --
> PLEASE NOTE NEW EMAIL ADDRESS BELOW
> *****************************************************
> James Wilfred Skerritt
> ()() Visiting Scientist (Genetics Unit)
> (..) Landcorp Farming Limited
> /\/\ Hochstetter House,
> ( ii ) Cnr Rostrevor & Clifton Streets
> \ /...... Hamilton, New Zealand
> E-mail: skerrittj@landcorp.co.nz (prefered)
> landhmf@wave.co.nz (at home)
> Homepage (www): http://www.wave.co.nz/pages/landhmf/
> snail: Landcorp Farming Ltd
> PO Box 1235, _--_|\
> Hamilton,New Zealand / \
> Home: 43B Hibiscus Avenue, \_.--._/ /|
> Hamilton,New Zealand v \|<=Hamilton
> phone: 07 838 0993 (work) /|`
> 07 838 1225 (home) \/
> fax: 07 838 0636 (ISD:64-7)
> *****************************************************
> Augustine says (De Doctr. Christ. i, 32), "Because He is good, we are."
>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD email: Arthur.Gilmour@agric.nsw.gov.au
Senior 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 currently free by anonymous ftp from pub/aar on ftp.res.bbsrc.ac.uk
Point your web browser at ftp://ftp.res.bbsrc.ac.uk/pub/aar/
in the IACR-Rothamsted information system http://www.res.bbsrc.ac.uk/
To join the asreml discussion list, send the message
subscribe
to asreml-request@chiswick.anprod.CSIRO.au
The address for messages to the list is asreml@chiswick.anprod.CSIRO.au
<> <> <> <> <> <> <>
"Seek first the kingdom of God and His righteousness" Jesus; Matthew 6: 33
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>