Dear all,
I'm trying to implement Hausman et al (1998) misclassification error for binary choice model with panel data. ml check give me an error : Two different coefficient vectors resulted in equal log likelihood values r(9). I don't know how I can fix it. Do you have any suggestions? Many thanks and best regards,
I'm trying to implement Hausman et al (1998) misclassification error for binary choice model with panel data. ml check give me an error : Two different coefficient vectors resulted in equal log likelihood values r(9). I don't know how I can fix it. Do you have any suggestions? Many thanks and best regards,
Code:
program drop _all program define mrpr version 16.1 args todo b lnfj tempvar mylnf xb alpha0 alpha1 last mleval `xb'=`b', eq(1) mleval `alpha0'=`b', eq(2) scalar mleval `alpha1'=`b', eq(3) scalar scalar `alpha0'=invlogit(`alpha0')/2 scalar `alpha1'=invlogit(`alpha1')/2 local by $My_panel sort `by' by `by' : gen `last'=(_n==_N) gen double `mylnf'=0 replace `mylnf'= $ML_y1 *ln(`alpha0'+(1-`alpha0'-`alpha1')*normal(`xb')) +(1-$ML_y1 )*ln((1-`alpha0')-(1-`alpha0'-`alpha1')*normal(`xb')) if `last'==1 mlsum `lnfj'=`mylnf' *if (`todo'==0 |`lnfj'>=.) exit end global My_panel personid ml model d0 mrpr (xb: yvar= `covariates' ) /alpha0 /alpha .ml check Test 1: Calling mrpr to check if it computes log likelihood and does not alter coefficient vector... Passed. Test 2: Calling mrpr again to check if the same log likelihood value is returned... Passed. Test 3: Calling mrpr to check if 1st derivatives are computed... test not relevant for type d0 evaluators. Test 4: Calling mrpr again to check if the same 1st derivatives are returned... test not relevant for type d0 evaluators. Test 5: Calling mrpr to check if 2nd derivatives are computed... test not relevant for type d0 evaluators. Test 6: Calling mrpr again to check if the same 2nd derivatives are returned... test not relevant for type d0 evaluators. ------------------------------------------------------------------------------ Searching for alternate values for the coefficient vector to verify that mrpr returns different results when fed a different coefficient vector: Searching... initial: log likelihood = -<inf> (could not be evaluated) searching for feasible values + feasible: log likelihood = -39456.325 improving initial values ++++...... improve: log likelihood = -37414.699 continuing with tests... ------------------------------------------------------------------------------ Test 7: Calling mrpr to check log likelihood at the new values... FAILED. Two different coefficient vectors resulted in equal log likelihood values of -37414.699. This does not prove there is a problem, but it suggests it. two coefficient vectors: xb: xb: xb: xb: xb: xb: /: /: did sexe mdefipc agegroup educgroup _cons alpha0 alpha1 r1 0 0 0 0 0 67.07219 69.31779 35.76148 r2 0 0 0 0 0 0 0 0 r(9);