Re: problems converging

With 100 clones it is hard to separate additive and non-additive effects,
and if you try to then you should probably include family as well. The
ratios of the variances are likely to be very variable between sites
however.

The model fitted will depend on the assumptions you make about variances and
correlations, and what you are trying to find out. I prefer to start with a
model that makes as few assumptions as possible. You can then test
different hypotheses by constraining the model.

d10 d18 ~ Trait Trait.site Trait.site.rep , #Trait.rep seems a strange
thing to fit
!r Trait.site.clone Trait.site.ide(clone) # Trait.rep.site with only 5 reps,
hard pressed to justify calling it random, except if wildly unbalanced. What
about plots?
2 2 2 # there are 2 sites, so it is almost assured that they will have
separate error variances.
N1 0 IDEN # N1 is the number of records on site 1
Trait 0 US 0.65 0.4 0.5698 !GU #starting covariance should not be so high as
0.9
N2 0 IDEN # N2 is the number of records on site 1
Trait 0 US 0.65 0.4 0.5698 !GU

Trait.site.clone 2 # fit full correlation matrix between sites and traits
Trait.site 0 CORR !+10 !GP
0.9
0.9 0.9
0.9 0.9 0.9
0.8 0.1213 0.8 0.213 #substitute better starting values
clone 0 AINV

Trait.site.ide(clone) 2 # fit full correlation matrix between sites and
traits
Trait.site 0 CORR !+10 !GP
0.9
0.9 0.9
0.9 0.9 0.9
0.8 0.1213 0.8 0.213 #substitute better starting values
ide(clone) 0 IDEN

Greg Dutkowski
Geneticist
PlantPlan Genetics Pty Ltd
School of Plant Science
Locked Bag 55 HOBART 7001 Tasmania, Australia
+61-3-62262213 +61-3-62262698 (fax) +61-408-238495 (mob)

-----Original Message-----
From: ASReml users discussion group [mailto:ASREML-L_at_AGRIC.NSW.GOV.AU] On
Behalf Of Heidi.Dungey_at_ENSISJV.COM
Sent: Tuesday, 1 November 2005 9:49 AM
To: ASREML-L_at_AGRIC.NSW.GOV.AU
Subject: problems converging

----- Forwarded by Heidi Dungey/ANZSC/NZ on 01/11/2005 11:47 -----

Dear Arthur,
I'm having problems getting convergence with some clonal data. It is across
2 sites and ~ 100 clones with at least 2 ramets per clone for most clones at
each site (although there may be some with only 1).

I can't get this to converge, even when I plug in across-site estimated
starting values.

Am I trying to get too much from this data set? Your help would be
appreciated.

cheers
Heidi Dungey
Ensis - the joint forces of CSIRO and SCION
Te Papa Tipu Innovation Park, 49 Sala Street
Private Bag 3020, ROTORUA, New Zealand
Phone 64-7-343 5629
Fax 64-7-348 0952
Home 64-7-362 4117
www.ensisjv.com
heidi.dungey_at_ensisjv.com

clonal data 2 sites dbh
 clone !P !I
 female 19 !I
 male 19 !I
 site 2 !A
 rep 6
 sets 10 !A
 tag 513 !A
 ramet 194 !I
 H1 H2 H3 H4 H7 HTM d4 d7 d10 d18 d19 st4 st7 st10 st18 br4 br7 br10 nra07
LN BS I1 I2 m_IML all_m_k_RAMS_CLONE.dat !repeat !make
all_m_k_RAMS_CLONE.dat !maxit 999 !step 0.1 d10 d18 ~ Trait Trait.site
Trait.rep !r Trait.clone Trait.ide(clone) Trait.clone.site
Trait.ide(clone).site Trait.rep.site !f mv 1 2 3 0 Trait 0 US 0.65 0.99
0.5698 !GU # Trait.clone 2 Trait 0 CORR .1405 0.8 0.1213 !GP clone 0 0 #
Trait.ide(clone) 2
Trait 0 CORR 0.2476 0.73 0.3108 !GU
ide(clone) 0 0
#
Trait.clone.site 2
Trait 0 CORR 0.2 0.8 0.01 !GP
clone.site 0 0
#
Trait.ide(clone).site 2
Trait 0 US 0.00005 0.5 0.0000134 !GU
ide(clone).site 0 0
#
Trait.rep.site 2
Trait 0 CORR .57 0.3 2.3 !GP
rep.site 0 0
#

 ASReml [ 4 Jan 2001]
 01 Nov 2005 11:26:41.555 256.00 Mbyte MSWIN asTBGallcorr_CL_ped_acsites

************************************************************
 A valid license to run ASReml was not found
 You are not permitted to use ASReml without a valid license except for a
trial period of up to 30 calendar days
************************************************************

 *****************************************************
 * ASReml 2000 Jun Default spline has max of 50 knots*
 * Sep Predict statement initiated *
 *********************************************** ARG *
  female 19 !I
  male 19 !I
  site 2 !A
  sets 10 !A
  tag 513 !A
  ramet 194 !I
  Reading pedigree file all_m_k_RAMS_CLONE.dat : skipping 0 lines
  Pedigree check: 850055 K 1 e 290 8800 occurs as Sire and Dam: Line 565 3
2 72 82
  Pedigree check: 875046 875047 K 1 b 4 occurs as Sire and Dam: Line 1533
2 3 13 216
 PEDIGREE [all_m_k_RAMS_CLONE.dat ] has 230 identities, 638 Non
zero elements
 QUALIFIERS: !MAXIT 999 !STEP 0.1
  Reading all_m_k_RAMS_CLONE.dat FREE FORMAT skipping 0 lines Bivariate
analysis of d10 and d18
 Using 1865 records of 1865 read
  Model term Size Type COL Minimum Mean Maximum #zero
#miss
   1 clone 230 Direct 1 1.000 108.0 194.0 0
0
   2 female 19 Factor 2 1 9.5303 19 0
0
   3 male 19 Factor 3 1 9.5260 19 0
0
   4 site 2 Factor 4 1 1.5228 2 0
0
  Warning: Fewer levels found in rep than specified
   5 rep 6 Factor 5 1 2.7469 5 0
0
   6 sets 10 Factor 6 1 4.9759 10 0
0
   7 tag 513 Factor 7 1 111.1700 513 0
0
   8 ramet 194 Factor 8 1 108.0032 194 0
0
   9 H1 1 Covariat 9 -3.891 0.2874E-05 1.747 0
334
  10 H2 1 Covariat 10 -4.253 0.4567E-05 4.900 0
4
  11 H3 1 Covariat 11 -4.186 0.3104E-05 4.609 0
61
  12 H4 1 Covariat 12 -4.219 0.5583E-05 3.941 7
74
  13 H7 1 Covariat 13 -3.692 -0.2546E-05 3.608 0
176
  14 HTM 1 Covariat 14 -3.123 0.1170E-05 7.665 0
1181
  15 d4 1 Covariat 15 -4.483 -0.6040E-05 3.003 1
971
  16 d7 1 Covariat 16 -3.129 -0.2614E-05 3.074 0
258
  17 d10 1 Variate 17 -3.725 0.2419E-05 4.035 0
294
  18 d18 1 Variate 18 -2.664 -0.3567E-05 8.524 0
1052
  19 d19 1 Covariat 19 -2.732 0.1726E-05 3.057 0
996
  20 st4 1 Covariat 20 -2.664 -0.6158E-06 2.822 0
1053
  21 st7 1 Covariat 21 -2.652 0.4922E-05 1.784 1
971
  22 st10 1 Covariat 22 -3.543 0.7778E-05 2.080 0
258
  23 st18 1 Covariat 23 -3.011 0.8640E-05 1.877 0
291
  24 br4 1 Covariat 24 -2.133 -0.1398E-04 2.206 1
971
  25 br7 1 Covariat 25 -2.572 -0.2676E-05 1.413 0
258
  26 br10 1 Covariat 26 -2.167 0.5587E-05 1.570 0
290
  27 nra07 1 Covariat 27 -4.432 0.8836E-05 2.083 0
258
  28 LN 1 Covariat 28-0.7346 0.9770E-05 6.847 0
995
  29 BS 1 Covariat 29-0.9058 -0.5409E-05 4.423 0
996
  30 I1 1 Covariat 30 -3.204 0.1483E-05 3.126 0
1056
  31 I2 1 Covariat 31 -3.216 -0.2974E-05 4.633 0
1058
  32 m_IML 1 Covariat 32 -3.439 0.1978E-05 3.168 0
1056
  33 Trait 2 Traits/Variat
  34 Trait.site 4 Interaction 33 Trait : 2 4 site :
2
  35 Trait.rep 12 Interaction 33 Trait : 2 5 rep :
6
  36 Trait.clone 460 Interaction 33 Trait : 2 1 clone :
230
  37 ide(clone) 230 Identity 1 1 108.0032 194 0
0
  38 Trait.ide(cl 460 Interaction 33 Trait : 2 37 ide(clone) :
230
  39 clone.site 460 Interaction 1 clone : 230 4 site :
2
  40 Trait.clone. 920 Interaction 33 Trait : 2 39 clone.site :
460
  41 ide(clone).s 460 Interaction 37 ide(clon: 230 4 site :
2
  42 Trait.ide(cl 920 Interaction 33 Trait : 2 41 ide(clone).site:
460
  43 rep.site 12 Interaction 5 rep : 6 4 site :
2
  44 Trait.rep.si 24 Interaction 33 Trait : 2 43 rep.site :
12
  45 mv_estimates 1346 Missing value
   1865 identity
 Warning: This US structure is not positive definite
      2 UnStructure 0.6500 0.9900 0.5698
    3730 records assumed sorted 2 within 1865
      2 CORRelation 0.1405 0.8000 0.1213
    230 Ainverse
 Structure of Trait.clone has 460 levels defined
      2 CORRelation 0.2000 0.8000 0.0100
    460 identity
 Structure of Trait.clone. has 920 levels defined
      2 CORRelation 0.5700 0.3000 2.3000
     12 identity
 Structure of Trait.rep.si has 24 levels defined
  Forming 2768 equations: 18 dense
 Initial updates will be shrunk by factor 0.100
 NOTICE: 1350 (more) singularities,
 LogL=-1083.23 S2= 1.0000 3716 df
 LogL=-967.304 S2= 1.0000 3716 df
 LogL=-729.480 S2= 1.0000 3716 df
 LogL= 54.8429 S2= 1.0000 3716 df
 LogL=-466.053 S2= 1.0000 3716 df
 Logliklihood decreased to -466.05: Trying again with reduced updates
 LogL=-18.8946 S2= 1.0000 3716 df
 Logliklihood decreased to -18.89: Trying again with reduced updates
 LogL= 47.4049 S2= 1.0000 3716 df
 LogL= 37.2687 S2= 1.0000 3716 df
 LogL=-40.5678 S2= 1.0000 3716 df
 Logliklihood decreased to -40.57: Trying again with reduced updates
 LogL= 30.6232 S2= 1.0000 3716 df
 LogL= 21.7377 S2= 1.0000 3716 df
 LogL=-54.6022 S2= 1.0000 3716 df
 Fault 0 Convergence Failed
  Last line read was:
 asTBGallcorr_CL_ped_acsites all_m_k_RAMS_CLONE.dat
  Model specification: TERM LEVELS GAMMAS
 Trait.rep 12 0.000
 Trait.site 4 0.000
 Trait 2 0.000
 Trait.clone 460 0.000
 Trait.clone.site 920 0.000
 Trait.rep.site 24 0.000
 mv_estimates 1346 0.000
SECTIONS 3730 8 1
    TYPE 0 0 0 2003 2005 2007 0
  STRUCT 1865 0 0 0 0 1 0
               2 3 9 9 3 1 0
               2 3 5 12 1 1 9
             230 0 -2 0 0 1 0
               2 3 5 15 1 1 9
             460 0 0 0 0 1 0
               2 3 5 18 1 1 9
              12 0 0 0 0 1 0
   45 factors defined [max 500].
   20 variance parameters [max 600]. 2 special structures
 Final parameter values 1.0000 0.67016 0.99000
  0.56980 0.14050 0.62693 0.12130 0.27042 0.80000 0.11607
E-01
  0.43260 0.27638E-010.48323
  Last line read was:
    45 0 1865 3730 32000
 Finished: 01 Nov 2005 11:26:47.754 Convergence Failed

Heidi Dungey
Scientist, Genetics,
Ensis - the joint forces of CSIRO and SCION
Te Papa Tipu Innovation Park, 49 Sala Street
Private Bag 3020, ROTORUA, New Zealand
Phone 64-7-343 5629
Fax 64-7-348 0952
Home 64-7-362 4117
www.ensisjv.com
heidi.dungey_at_ensisjv.com
************************************************************
CONFIDENTIALITY NOTICE/DISCLAIMER:
The information in this transmission (which includes any attachments) is
confidential and may also be legally privileged. It is intended only for the
use of the individual or entity to which it is addressed. If you are not the
intended recipient, please be aware that it is prohibited to read, use,
disseminate, distribute or copy this transmission or any part of it. If you
have received this transmission in error, please notify us immediately and
destroy this transmission. Forest Research accepts no responsibility for
viruses or anything similar in this transmission nor does not accept
responsibility for any changes made to this transmission after sending.
Thank you.

---------------------------------------------------------
This email is confidential and may be legally privileged.
If received in error please destroy and immediately notify us.
Received on Mon Nov 01 2005 - 10:56:24 EST