On Mon, 19 Oct 2009, Dorian Litvine wrote: > My name is Dorian, I'm trying to estimate the data of a double bounded > dichotomous choice (contingent valuation) thanks to a maximum > likelihood estimation with a linear specification (From E. Flachaire). > But when I put the command, then GRETL gives several errors messages > as "missing values found". I think the model may not be specified for > linear. But it is strange because when I put initial values quite high > (5) then it converges but I don't know if results are "possible". May > someone help me? Here is the code > > genr alpha=3 > genr sigma=1 > logl=(1-CD1)*(1-CD2)*log(Pnn)+(1-CD1)*CD2*log(Pny)+CD1*(1-CD2)*log(Pyn)+CD1*CD2*log(Pyy) > series Pnn=1/(1+exp(-(BID2-alpha)/sigma)) > series > Pny=1/(1+exp(-(BID1-alpha)/sigma))-1/(1+exp(-(BID2-alpha)/sigma))+(BID2>BID1)*100 > series > Pyn=1/(1+exp(-(BID2-alpha)/sigma))-1/(1+exp(-(BID1-alpha)/sigma))+(BID2<BID1)*100 > series Pyy=1-1/(1+exp(-(BID2-alpha)/sigma)) > params alpha sigma > > And attached you have the results if I set the initial values at 5.
You need to tell us what your model actually is. You give a code fragment above, and this could be turned into a block requesting mle estimation by wrapping all but the first two lines in mle ... <your code here> end mle But the code does not match the results you attached: above we have only two parameters, alpha and sigma, while you report estimates of three: alpha0, alpha1 and sigma. Allin Cottrell
