Latent variable and panel data

[Lorena Torres]

Hello, 

I am doing a model with a latent variable with an error component normally distributed and I am using panel data, therefore, I am adding an error term in the utility of the choice that it is also normally distributed across individuals to account for serial correlation. 

I got estimates, but the problem is that the correlation coefficient for all the variables is infinite (1.8e+308). What can cause this type of estimation error?

This is the utility function where car is the latent variable previously defined.

INTER_INV = Beta('INTER_INV', 1, -1000 ,1000 ,0)

INTER = INTER_INV * bioDraws('INTER_STD', 'NORMAL')

# Función de utilidad

V = BETA_CAR * CAR + INTER_BASE +INTER + Intercept

Prob = ordered_logit(

continuous_value=V,

list_of_discrete_values=[1, 2, 3],

tau_parameter=tau1_V,

)

Prob_chosen = Elem(Prob, Relax)

condlike = (Prob_chosen * P_Car1 * P_Car2 * P_Car3)

#Panel data

condprobIndiv = PanelLikelihoodTrajectory(condlike)

loglike = log(MonteCarlo(condprobIndiv))