Re: samin estimators' errors

[Olaf Till]

On Tue, May 09, 2017 at 02:11:34PM -0700, Emrocar wrote:
> I implemented an optimization routine using samin to minimize a non-linear
> function with 7 free parameters. Now, the issue that I have is that I can't
> find the standard error associated to the estimators that samin gave me. Is
> there anybody who could please give me a hand?

'samin' is a stochastic optimizer, so it has no way to do it.

We have functions to return estimated statistics for residual based
optimization (as e.g. curve fitting), see residmin_stat.

For optimization of scalars, the optimizer has no way to guess the
standard deviation of the data, so you must provide a guess of this
yourself. Then, you can use the gradient and the Hessian at the
solution to compute your estimate of the standard deviation of the
estimated parameters. But currently none of our scalar nonlinear
optimizers can return the gradient and the Hessian at the solution,
sorry.

Olaf

-- 
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net

signature.asc
Description: Digital signature