EViews.com

I'm estimating some non linear equations and naturally am finding that results are often pretty sensitive to the initial parameter conditions in the param statement.

I wonder therefore if anyone knows how to take all the parameters (or a subset) and simply, say, draw their value from a Normal distribution, estimate the eqn and then save those intial condiitions and some suitable metric (like R^2 or the loglik). Then do this repetedly in a loop.

I guess an important consideration would be an error handler since some of the drawn parameter vector would lead to non convergent solutions.

Thanks in advance!

Typing: will create standard normal random numbers in the C vector, which is used for starting values.