OPTMODEL to maximize likelihood

jpunch — Tue, 27 Apr 2010 15:50:45 GMT

I am trying to maximize a likelihood for independent founders using OPTMODEL (I have already written the code to perform a simple grid search but want something quicker). I have a dataset QTp with only one variable (also called QTp). Is the following code valid?

proc optmodel;

set S;

number t{S};

var q GE 0 init 0.5;

var muaa init 0;

var muab init 0;

var mubb init 0;

var sigma>=0.000001 init 0.05;

max logl= sum{i in S} LOG(1/(SQRT(2*3.14)*sigma))+
sum{i in S} LOG(((1-q)**2)*(EXP(-((t-muaa)**2)/(2*((sigma)**2)))+ (2*q*(1-q))*EXP(-((t-muab)**2)/(2*((sigma)**2)))+
((q)**2)*EXP(-((t-mubb)**2)/(2*((sigma)**2))));

read data QTp into S=[_n_] t=QTp;

solve with nlpc / tech=cgr printfreq=1 maxiter=5000;

print q sigma muaa muab mubb;

quit;

Message was edited by: jpunch

I also want q LE 1 but can't put it in the text window. Thanks.

Message was edited by: jpunch Message was edited by: jpunch

Re: OPTMODEL to maximize likelihood

IoannisAkrotirianakis — Tue, 04 May 2010 16:02:32 GMT

There were several parentheses that were missing.
If you try the following code you should not get any syntax errors:

proc optmodel;

set S;

number t{S};

var q >= 0 <= 1 init 0.5;

var muaa init 0;

var muab init 0;

var mubb init 0;

var sigma>=0.000001 init 0.05;

max log= sum{i in S}
LOG(1/(SQRT(2*3.14)*sigma)
)
+
sum{i in S}
LOG(
((1-q)**2)*
(EXP(
-((t-muaa)**2)/(2*((sigma)**2)))
+
(2*q*(1-q))*EXP(
-((t-muab)**2)/(2*((sigma)**2))
)
+
((q)**2)*EXP(
-((t-mubb)**2)/(2*((sigma)**2))
)
)
);

read data QTp into S=[_n_] t=QTp;

solve with nlpc / tech=cgr printfreq=1 maxiter=5000;

print q sigma muaa muab mubb;

quit;

topic Re: OPTMODEL to maximize likelihood in Mathematical Optimization, Discrete-Event Simulation, and OR

OPTMODEL to maximize likelihood

Re: OPTMODEL to maximize likelihood