Dispersion parameter
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Dispersion parameter



Hi,
I am working with the fitting a Poisson repeatability model to ovulation
rate records.  The model is

or !POI !LOG ~ mu breed year age !r animal perm


With Asreml, this eventually sends both animal and perm variance
components to zero (tornd.asr).  
 Source                Model  terms     Gamma     Component    Comp/SE  
% C
 animal                  854    854  0.258454E-02  0.258454E-02   0.16
-28 P   
 perm                    358    358  0.888961E-07  0.888961E-07   0.00
-99 P   
 Variance                937    921   1.00000       1.00000       0.00  
0 F   


However, if I include the dispersion option (!DISP) I get reasonable
values (torn2.asr):

Source                Model  terms     Gamma     Component    Comp/SE  
% C
 animal                  854    854  0.421773E-01  0.702392E-02   1.38  
0 P   
 perm                    358    358  0.921755E-01  0.153503E-01   2.98  
0 P   
 Variance                937    921   1.00000      0.166533      17.95  
0 U   

I am hoping that would be able to explain this to me, especially the
intrepretation of the dispersion parameter in this context and providing
estimates of heritability and repeatability.

Thanks in advance,
Bruce
  ASREML [21 Mar 2000]  Generalized mixed models                                                
 Wed Apr 12 17:34:25 2000   8.00 Mbyte  Unix   torn2
 *****************************************************
 * ASREML 2000  Residuals now written to .yht file,  *
 * convergence monitoring of parameters to .res file *
 *********************************************** ARG *
 A-inverse retrieved from ainverse.bin
 PEDIGREE [ped.coded ] has      854 identities,    2823 Non zero elements
 QUALIFIERS: !NODISPLAY                                                  
  Reading asreml.data  FREE FORMAT skipping  0  lines
 Univariate analysis of or                  
 Using      937 records [of     948 read from     948 lines of asreml.data         ]
  Model term      Size Type    COL   Minimum    Mean      Maximum   #zero #miss
   1 animal        854 Direct    1   215.0     581.8      837.0         0     0
   2 perm          358 Factor    2      1   161.9733        358         0     0
   3 breed           6 Factor    3      1     2.3927          6         0     0
   4 year            8 Factor    4      1     4.4440          8         0     0
   5 age             4 Factor    5      1     2.5955          4         0     0
   6 or              1 Variate   6 0.5000      2.572      9.000         0     0
   7 nlb             1 Covariat  7  1.000      2.017      5.000         0   159
   8 mu              1 Constant Term
  Forming  1231  equations:  19  dense
 Initial updates will be shrunk by factor    0.010
 Distribution and link: Poisson; Log   Mu=exp(XB) V=Mu                      
 NOTICE:     3 (more) singularities,
 LogL= 640.120     S2=  1.0000        921 df   0.1000     0.1000      1.000    

 LogL= 627.491     S2=  1.0000        921 df   0.8906E-01 0.8890E-01  1.000    

 LogL= 708.005     S2=  1.0000        921 df   0.8906E-02 0.8890E-02  1.000    

 LogL= 714.419     S2=  1.0000        921 df   0.2936E-02 0.8890E-03  1.000    

 LogL= 714.424     S2=  1.0000        921 df   0.3638E-02 0.8890E-05  1.000    

 Final parameter values                       0.25845E-020.88896E-07 1.0000    

 Deviance from GLM fit                    -204.01
 Variance heterogenity factor [Deviance/DF]     0.22

 Source                Model  terms     Gamma     Component    Comp/SE   % C
 animal                  854    854  0.258454E-02  0.258454E-02   0.16 -28 P   
 perm                    358    358  0.888961E-07  0.888961E-07   0.00 -99 P   
 Variance                937    921   1.00000       1.00000       0.00   0 F   

 Analysis of Variance       DF     F-incr      F-adj  StndErrDiff
   8 mu                      1    1717.95      99.84
   3 breed                   5      34.97      31.71  0.1037    
   4 year                    7       1.27       1.11  0.8591E-01
   5 age                     3       3.69       3.69  0.6120E-01

                     Solution       Standard Error    T-value     T-prev
   5 age                 
                    4  -0.378294E-01   0.614788E-01     -0.62
                    3  -0.828053E-01   0.594600E-01     -1.39     -0.71
                    2  -0.192746       0.604171E-01     -3.19     -1.88
   4 year                
                 1993   0.489248E-02   0.883469E-01      0.06
                 1994  -0.451166E-01   0.856120E-01     -0.53     -0.59
                 1995  -0.419165E-01   0.848745E-01     -0.49      0.04
                 1996   0.800418E-01   0.902823E-01      0.89      1.55
                 1997   0.680066E-01   0.891103E-01      0.76     -0.15
                 1998  -0.691629E-01   0.939238E-01     -0.74     -1.73
                 1999   0.925343E-01   0.985045E-01      0.94      1.83
   3 breed               
 6                      0.507577       0.651190E-01      7.79
 3                     -0.850997E-01   0.106344         -0.80     -5.36
 4                      0.580041       0.851034E-01      6.82      4.93
 5                      0.499146       0.561617E-01      8.89     -0.89
 2                     -0.109632       0.101933         -1.08     -5.76
   8 mu                  
                   19   0.825678       0.826323E-01      9.99
   1 animal                 854 effects fitted
   2 perm                   358 effects fitted
 5  possible outliers: see .res file
 Finished: Wed Apr 12 17:34:30 2000   WARNING: LogL Converged; Parameters Not Converged           
  ASREML [21 Mar 2000]  Generalized mixed models                                                
 Wed Apr 12 17:30:55 2000   8.00 Mbyte  Unix   torn2
 *****************************************************
 * ASREML 2000  Residuals now written to .yht file,  *
 * convergence monitoring of parameters to .res file *
 *********************************************** ARG *
 A-inverse retrieved from ainverse.bin
 PEDIGREE [ped.coded ] has      854 identities,    2823 Non zero elements
 QUALIFIERS: !NODISPLAY                                                  
  Reading asreml.data  FREE FORMAT skipping  0  lines
 Univariate analysis of or                  
 Using      937 records [of     948 read from     948 lines of asreml.data         ]
  Model term      Size Type    COL   Minimum    Mean      Maximum   #zero #miss
   1 animal        854 Direct    1   215.0     581.8      837.0         0     0
   2 perm          358 Factor    2      1   161.9733        358         0     0
   3 breed           6 Factor    3      1     2.3927          6         0     0
   4 year            8 Factor    4      1     4.4440          8         0     0
   5 age             4 Factor    5      1     2.5955          4         0     0
   6 or              1 Variate   6 0.5000      2.572      9.000         0     0
   7 nlb             1 Covariat  7  1.000      2.017      5.000         0   159
   8 mu              1 Constant Term
  Forming  1231  equations:  19  dense
 Initial updates will be shrunk by factor    0.010
 Distribution and link: Poisson; Log   Mu=exp(XB) V=Mu                      
 NOTICE:     3 (more) singularities,
 LogL= 1118.23     S2= 0.15158        921 df   0.1000     0.1000      1.000    

 LogL= 1081.26     S2= 0.15803        921 df   0.9933E-01 0.1000      1.000    

 LogL= 1081.86     S2= 0.15893        921 df   0.9175E-01 0.9909E-01  1.000    

 LogL= 1083.29     S2= 0.16162        921 df   0.7175E-01 0.9633E-01  1.000    

 LogL= 1084.29     S2= 0.16476        921 df   0.5291E-01 0.9277E-01  1.000    

 LogL= 1084.58     S2= 0.16672        921 df   0.4236E-01 0.9107E-01  1.000    

 LogL= 1084.55     S2= 0.16651        921 df   0.4218E-01 0.9228E-01  1.000    

 LogL= 1084.55     S2= 0.16653        921 df   0.4218E-01 0.9216E-01  1.000    

 Final parameter values                       0.42177E-010.92176E-01 1.0000    

 Deviance from GLM fit                    -133.67
 Variance heterogenity factor [Deviance/DF]     0.15

 Source                Model  terms     Gamma     Component    Comp/SE   % C
 animal                  854    854  0.421773E-01  0.702392E-02   1.38   0 P   
 perm                    358    358  0.921755E-01  0.153503E-01   2.98   0 P   
 Variance                937    921   1.00000      0.166533      17.95   0 U   

 Analysis of Variance       DF     F-incr      F-adj  StndErrDiff
   8 mu                      1    2318.23     360.66
   3 breed                   5      90.78      79.34  0.5700E-01
   4 year                    7       8.66       6.69  0.3951E-01
   5 age                     3      16.70      16.70  0.2773E-01

                     Solution       Standard Error    T-value     T-prev
   5 age                 
                    4  -0.423435E-01   0.271356E-01     -1.56
                    3  -0.870193E-01   0.281661E-01     -3.09     -1.65
                    2  -0.202955       0.305604E-01     -6.64     -4.66
   4 year                
                 1993   0.292374E-02   0.373854E-01      0.08
                 1994  -0.581240E-01   0.373076E-01     -1.56     -1.74
                 1995  -0.710062E-01   0.383299E-01     -1.85     -0.40
                 1996   0.461848E-01   0.421068E-01      1.10      3.50
                 1997   0.507821E-01   0.438866E-01      1.16      0.14
                 1998  -0.102984       0.474361E-01     -2.17     -4.55
                 1999   0.733845E-01   0.512010E-01      1.43      4.70
   3 breed               
 6                      0.494528       0.400377E-01     12.35
 3                     -0.487666E-01   0.530696E-01     -0.92     -9.88
 4                      0.564815       0.568147E-01      9.94      8.17
 5                      0.504536       0.374205E-01     13.48     -1.04
 2                     -0.101050       0.517223E-01     -1.95    -11.13
   8 mu                  
                   19   0.832114       0.438158E-01     18.99
   1 animal                 854 effects fitted
   2 perm                   358 effects fitted
 4  possible outliers: see .res file
 Finished: Wed Apr 12 17:31:02 2000   LogL Converged                                              

torn2.as