unbalanced repeated measures

Dear group members,

I’m trying to get a grip on a dataset where I measured calling effort in
males at specific ages during their adult life (5, 12, 19 days post
eclosion, and from then on every 4th day; maximum number of measurements
is 13).

My datafile contains my repeated measures as a single variable (‘call’),
and the age at which the measure was taken as another variable
(‘callage’).I measured 81 animals, and there are 631 missing values (out
of 1053 theoretical measurements from 81*13).

In the example below I want to start with a basic model (‘call’ as the
dependent variable, and ‘callage’ as the age at which the measurement
was taken).
Later on I want to include covariates (size at eclosion, ambient
temperature at time of measurement, eclosing date) in the model.

The model given below runs and the LL converges, but I’m not really sure
whether I got the model right, especially the !ASMV qualifier: when I
put ‘13’ (=number of levels) behind ‘!ASMV’ in the variance header, the
job doesn’t run.

The second point is that I’m still struggling to try out different
covariance structures, eg. antedependence. As in the model below, the
initial values are not given correctly. ASREML fixes them to 0.001 when
running the job (see .asr output below), but when I put ‘0.001’ between
‘AR1’ and ‘!GP’, the LL doesn’t converge. I don’t know how to correctly
specify them, and would appreciate any hint/help on that matter.

Cheers,
Felix

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
::::
JOBFILE:

Analysis of longitudinal data for males
         ind !I
         eclosed !DATE
         size
         temp
         callage !I #13 levels of age at which CE was measured
(max=59)
         call
males-readyFORasremlreduced.txt !SKIP1
call !POISSON !DISP ~ callage !r callage.ind !f mv
1 2 1 !ASMV
81
callage 0 AR1 !GP
13*0.1
callage.ind 2
callage 0 AR1 !GP
13*0.1
ind

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::::::
.ASR OUTPUT:

SReml 2.00a [01 Jul 2006] Analysis of longitudinal data for males
     Build: h [ 9 Sep 2006] 32 bit
 10 Oct 2006 12:44:12.255 32.00 Mbyte Windows
in\correl-longitudinal-reduced
 Running under 30 day Demonstration License
 ***********************************************************
 * SYNTAX change: A/B now means A A.B *
 * *
 * Contact support_at_asreml.co.uk for licensing and support *
 ***************************************************** ARG *
 Folder: D:\academic\PhD\general\progs\ASREML\ASREML2\bin
          ind !I
          eclosed !DATE
          callage !I
 QUALIFIERS: !SKIP1
 Reading males-readyFORasremlreduced.txt FREE FORMAT skipping 1
lines

 Univariate analysis of call
 Using 1053 records of 1053 read

  Model term Size #miss #zero MinNon0 Mean
MaxNon0
   1 ind 81 0 0 1 41.0000
81
   2 eclosed 0 0 0.3876E+05 0.3877E+05
0.3879E+05
 Warning: If eclosed is fitted as a covariate, it should be centred
first.
   3 size 0 0 5.870 6.678
7.500
   4 temp 0 0 16.20 21.43
24.60
   5 callage 13 0 0 1 7.0000
13
   6 call Variate 631 207 1.000 5.005
24.00
   7 callage.ind 1053 5 callage : 13 1 ind
: 81
   8 mv_estimates 631
     81 identity

 Warning: Invalid initial value for AR=AutoReg parameter 1 has been
changed
          to 0.001 and fixed. The variance of the data is around
3.656
     13 AR=AutoReg 0.0010
    1053 records assumed pre-sorted 13 within 81

 Warning: Invalid initial value for AR=AutoReg parameter 1 has been
changed
          to 0.001 and fixed. The variance of the data is around
3.656
     13 AR=AutoReg 0.0010
     81 identity
 Structure for callage.ind has 1053 levels defined
 Forming 1697 equations: 13 dense.
 Initial updates will be shrunk by factor 0.010
 Distribution and link: Poisson; Log Mu=exp(XB) V=Mu

 Warning: The LogL value is unsuitable for comparing GLM models
   1 LogL=-709.968 S2= 1.5777 409 df 1.000 0.1000E-02
0.1000E-02

   2 LogL=-628.968 S2= 1.7628 409 df 1.000 0.1000E-02
0.1000E-02

   3 LogL=-624.167 S2= 1.7794 409 df 1.000 0.1000E-02
0.1000E-02

   4 LogL=-624.163 S2= 1.7795 409 df 1.000 0.1000E-02
0.1000E-02

   5 LogL=-624.163 S2= 1.7795 409 df 1.000 0.1000E-02
0.1000E-02

   6 LogL=-624.163 S2= 1.7795 409 df 1.000 0.1000E-02
0.1000E-02

 Final parameter values 1.0000
0.10000E-020.10000E-02

 Deviance from GLM fit 409 404.96
 Variance heterogenity factor [Deviance/DF] 0.99

 Source Model terms Gamma Component Comp/SE
% C
 Variance 1053 409 1.00000 1.77945 14.30
0 U
 Residual AR=AutoR 13 0.100000E-02 0.100000E-02 0.00
0 F
 callage.ind AR=AutoR 13 0.100000E-02 0.100000E-02 0.00
0 F

 Analysis of Variance NumDF DenDF F_inc
Prob
   5 callage 13 409.0 12.70
<.001
 Notice: The DenDF values are calculated ignoring
fixed/boundary/singular
             variance parameters using algebraic derivatives.

 Warning: This Analysis of Variance based on the working variable is not
          equivalent to the Analysis of Deviance. Standard errors are
scaled
          by the variance of the working variable, not the residual
deviance.
   8 mv_estimates 631 effects fitted
   7 callage.ind 1053 effects fitted ( 401
are zero)
 SLOPES FOR LOG(ABS(RES)) on LOG(PV) for Section 1
   2.88
        63 possible outliers: in section 1 (see .res file)
 Finished: 10 Oct 2006 12:44:19.215 LogL Converged
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
:::::::::

______________________________________________________________
::Felix Zajitschek
::School of Biological, Earth and Environmental Sciences (BEES)
::University of New South Wales (UNSW) NSW 2052 - Australia
::Labtel +61 (0)2 9385 2124
::Fax +61 (0)2 9385 1558
::eMail felix.zajitschek_at_student.unsw.edu.au
::www.bees.unsw.edu.au/school/researchstudents/zajitschekfelix.html
Received on Sun Oct 10 2006 - 13:37:54 EST