Local maximum of LogL
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Local maximum of LogL



Dear Arthur, Dear All,

When fitting a univariate model I get the following problem: The LogL
function converge at a maximum value. Sometimes  the value decreases in
every iteration and sometimes it first increases and then decreases, but
the final value is not the maximum. Changing the initial values or using
!GU option makes the LogL to increase in every itration, but the final
value is the same as in prior run.

So I think ASREML has found a local maximum, but not the global maximum.
Is this right ? Any sugestion on how to solve this problem ?

Hermann

PS: Following are the LogL values so you can understan my questions.

LogL=-1177.45     S2=  5.1233        802 df   0.1000     0.1000
1.000

 LogL=-1177.60     S2=  4.9262        802 df   0.1277     0.1299
1.000

 LogL=-1178.34     S2=  4.6714        802 df   0.1730     0.1699
1.000

 LogL=-1179.18     S2=  4.4919        802 df   0.2145     0.1949
1.000

 LogL=-1179.47     S2=  4.4425        802 df   0.2288     0.1993
1.000

 LogL=-1179.50     S2=  4.4371        802 df   0.2305     0.1996
1.000

 LogL=-1179.50     S2=  4.4366        802 df   0.2306     0.1996
1.000

----------



--
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml