Re: Maximum likelihood and Mathematica

*To*: mathgroup at smc.vnet.net*Subject*: [mg85312] Re: Maximum likelihood and Mathematica*From*: snowrain <anne_xuetulip at hotmail.com>*Date*: Tue, 5 Feb 2008 19:45:13 -0500 (EST)

The error info I got from estimating unconcerntrated loglikelihood function is as below: NMaximize::nrnum: The function value 377046.+157.08 \[ImaginaryI] is \ not a real number at {s11,s12,s13,s22,s23,s33,Subscript[b, \ 11],Subscript[b, 111],Subscript[b, 112],Subscript[b, 113],<<18>>} = \ {0.304936,0.266141,0.365626,0.132704,0.870404,0.952376,-0.523097,0.\ 275125,-0.797803,0.291049,<<18>>}. >> How can now check where does the imaginary number come from? I was thinking to put the estimated parameter values back to the loglikelihood function. But Mathematica does not give me a complete list of of estimates for parameters (18 of them were not shown in the output as above). How to check then? I have used options for NMaximize to check the procedures of estimation. Should I change AccuracyGoal and WorkingPrecision? I am not sure what changes these two options are gonna cause. thanks a lot !