Hi, I am writing some R code that fits a transfer function to a particular time series, using the optim() function (BFGS method) to minimize the sum-of-squared errors (SSE). I wish to estimate the standard errors of the parameter estimates returned by optim. Am I right in thinking that I can use the following std.err <- sqrt(diag(np*fit$value/(M - 2) * solve(fit$hessian))) where np are the number of parameters, fit$value is the SSE of the fitted model, M is the number of observations and fit$hessian is the hessian matrix returned by the optim() function? Kind regards, Sam. ------------------------------------- Samuel E. Kemp School of Computing University of Glamorgan South Wales UK [[alternative HTML version deleted]]