family variance changes
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
family variance changes
Dear ASREML users
I am somewhat puzzled by the extreme increase in family variance shown
below. The first analysis was for 1117 records, from 20 families. Then this
was rerun, with the addition of 2 families (& only 118 more records),
giving 1235 records and a total of 22 families. The family variance
increased from 0.682415E-04 to 18.7069. The incomplete block (lblk)
variance changed from 0.527994 to 0.693685E-04. Note that the gamma value
for family in the first analysis is the same as the gamma for incomplete
block in the second. Can anyone (Arthur? ) shed some light on these results?
Also, I have been trying to run a 5 trait multivariate, but cannot get
convergence. The first 4 traits work OK, but when I add the next (even
using results from the univariate for the starting values for the
variances, but stil guessing at covariances betwwen the fifth trait and the
other 4) I have problems. Any ideas? Or am I just pushing things a bit far?
Is doing 10 bivariates a safer (but more tedious) option?
Thanks in advance
Sue
ASREML [23 Oct 1998] OP trials -
12/08/98 17:49:40.60 8.00 Mbyte 3944a
! rerun with 4799 records Tues 8/12/98, lblk random
! families h6 & h15 excluded (because ramet collections) and
! no controls (entr<23)
QUALIFIERS: !skip 1 !filter 1 !select 4
Reading 394.csv FREE FORMAT skipping 1 lines
Univariate analysis of db8
Using 1117 records [of 4799 read from 4799 lines of 394.csv
]
Model term Size Type COL Minimum Mean Maximum #zero
#miss
1 site 4 Factor 1 4 4.0000 4 0
0
2 expt 4 Factor 2 4 4.0000 4 0
0
3 family 22 Factor 3 1 10.3769 20 0
0
4 blk 5 Factor 4 1 3.4691 6 0
0
5 lblk 6 Factor 5 1 2.9409 5 0
0
6 select 2 Factor 6 1 1.5944 2 0
0
7 ht1 1 Covariat 7 0.0000 0.0000 0.0000 0
1117
8 ht4 1 Covariat 8 0.0000 0.0000 0.0000 0
1117
9 db8 1 Variate 9 21.00 145.8 216.0 0
0
10 ht8 1 Covariat 10 1.000 118.2 168.0 0
0
11 st8 1 Covariat 11 1.000 1.324 4.000 0
1
12 db12 1 Covariat 12 27.00 182.9 276.0 0
10
13 ht12 1 Covariat 13 106.0 180.9 218.0 0
727
14 vs_il12 1 Covariat 14 1.000 2.490 6.000 0
40
15 nw12 1 Covariat 15 1.000 7.376 14.00 0
566
16 st12 1 Covariat 16 1.000 3.216 7.000 0
24
17 mu 1 Constant Term
18 family.lblk 132 Interaction 3 family: 22 5 lblk : 6
Forming 162 equations: 2 dense
Initial updates will be shrunk by factor 0.548
NOTICE: 48 (more) singularities,
LogL=-4231.79 S2= 658.21 1116 df 0.10000 0.10000 0.10000
1.0000
LogL=-4216.91 S2= 675.36 1116 df
0.10000E-010.10000E-010.35135E-01
1.0000
LogL=-4214.84 S2= 682.10 1116 df
0.10000E-020.10000E-020.25398E-01
1.0000
LogL=-4214.75 S2= 682.41 1116 df
0.10000E-030.73854E-030.25550E-01
1.0000
LogL=-4214.74 S2= 682.42 1116 df
0.10000E-040.77500E-030.25588E-01
1.0000
Final parameter values
0.10000E-060.77371E-030.25598E-01
1.0000
Source Model terms Gamma Component Compnt/StndErr
family 22 20 0.100000E-06 0.682415E-04 0.00 P
lblk 6 5 0.773714E-03 0.527994 0.15 P
family.lblk 132 87 0.255979E-01 17.4684 1.28 P
Variance 1117 1116 1.00000 682.415 22.84 P
Solution Standard Error T-value T-prev
17 mu 1 21889.55 21889.55 [DF
F_inc F_all]
1 145.821 0.985601 147.95
3 family 20 effects fitted
5 lblk 5 effects fitted
18 family.lblk 87 effects fitted
SLOPES FOR LOG(ABS(RES)) on LOG(PV) for Section 1
1.10
17 possible outliers: see .res file
Finished: 17:49:45.05 LogL Converged
ASREML [23 Oct 1998] OP trials -
12/08/98 19:38:29.70 8.00 Mbyte 3944b
! rerun with 5279 records Tues 8/12/98, lblk random
! families h6 & h15 NOT excluded, still no controls (entr<23)
QUALIFIERS: !skip 1 !filter 1 !select 4
Reading 394.csv FREE FORMAT skipping 1 lines
Univariate analysis of db8
Using 1235 records [of 5279 read from 5279 lines of 394.csv
]
Model term Size Type COL Minimum Mean Maximum #zero
#miss
1 site 4 Factor 1 4 4.0000 4 0
0
2 expt 4 Factor 2 4 4.0000 4 0
0
3 family 22 Factor 3 1 11.3741 22 0
0
4 blk 5 Factor 4 1 3.4680 6 0
0
5 lblk 6 Factor 5 1 2.9514 5 0
0
6 select 2 Factor 6 1 1.5935 2 0
0
7 ht1 1 Covariat 7 0.0000 0.0000 0.0000 0
1235
8 ht4 1 Covariat 8 0.0000 0.0000 0.0000 0
1235
9 db8 1 Variate 9 15.00 147.2 216.0 0
0
10 ht8 1 Covariat 10 1.000 118.8 168.0 0
0
11 st8 1 Covariat 11 1.000 1.387 4.000 0
1
12 db12 1 Covariat 12 17.00 184.5 276.0 0
10
13 ht12 1 Covariat 13 106.0 181.2 218.0 0
806
14 vs_il12 1 Covariat 14 1.000 2.464 6.000 0
42
15 nw12 1 Covariat 15 1.000 7.511 14.00 0
625
16 st12 1 Covariat 16 1.000 3.346 8.000 0
26
17 mu 1 Constant Term
18 family.lblk 132 Interaction 3 family: 22 5 lblk : 6
Forming 162 equations: 2 dense
Initial updates will be shrunk by factor 0.548
NOTICE: 36 (more) singularities,
LogL=-4686.05 S2= 665.74 1234 df 0.10000 0.10000 0.10000
1.0000
LogL=-4675.24 S2= 693.53 1234 df
0.10000E-010.10000E-010.25305E-01
1.0000
LogL=-4673.61 S2= 694.83 1234 df
0.21936E-010.10000E-020.14884E-01
1.0000
LogL=-4673.52 S2= 693.89 1234 df
0.26510E-010.10000E-030.14939E-01
1.0000
LogL=-4673.52 S2= 693.69 1234 df
0.26957E-010.10000E-040.15226E-01
1.0000
Final parameter values
0.26967E-010.10000E-060.15265E-01
1.0000
Source Model terms Gamma Component Compnt/StndErr
family 22 22 0.269674E-01 18.7069 1.72 P
lblk 6 5 0.100000E-06 0.693685E-04 0.00 P
family.lblk 132 97 0.152655E-01 10.5894 0.88 P
Variance 1235 1234 1.00000 693.685 23.89 P
Solution Standard Error T-value T-prev
17 mu 1 13886.94 13886.94 [DF
F_inc F_all]
1 147.263 1.24965 117.84
3 family 22 effects fitted
5 lblk 5 effects fitted
18 family.lblk 97 effects fitted
SLOPES FOR LOG(ABS(RES)) on LOG(PV) for Section 1
0.62
18 possible outliers: see .res file
Finished: 19:38:37.50 LogL Converged
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
HA & SF Jarvis
"Aramis" RSD 820
Dorodong via Penola
SA 5277
Phone/Fax: 03 55 838232 (+61 3 55 838232)
Email: aramis@dove.net.au
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml