Hello, I'm trying to estimate the beta distribution parameters (alpha, beta) using Proc NLMIXED. Because beta is not one of the distributions with built-in log likelihood, I used the general statement as follows: proc nlmixed data=test; parms a=5 b=5; *s=beta(a,b); s=gamma(a*b)/(gamma(a)*gamma(b)); ll=(1-a)*xbeta+(1-b)*(1-xbeta)-log(s); model xbeta~general(ll); run; I tried using the beta function directly and with gamma function. Both gave me the 'QUANEW optimization cannot be completed' error. The test data is just 1000 beta variates with alpha=beta=5. I even set the initial values to 5, but still got this error. Anne
... View more