[SciPy-dev] confidence intervals on multi-parameter minimisation
Wed Mar 21 18:55:17 CDT 2007
I've manually plugged in parameters with "mean + sd_beta" and "mean -
sd_beta" and the sd_beta values make sense from that point of view.
But I will follow up with your suggestion of a monte-carlo analysis
to look at the (residual**2).sum() vs "mean + random()"
On 22/03/2007, at 9:45 AM, Robert Kern wrote:
> Graeme O'Keefe wrote:
>> I've gone through the docstrings and test_odr.py, quite
>> straightforward, well packaged.
> Thank you!
>> I still run fmin_l_bfgs_b to bound the solution and then I use that
>> as a starting point.
>> Of course, now I know my model is really bad from the parameter
> Take those uncertainty estimates with a grain of salt. They are
> based on a
> linearization (quadraticization, really) of the loss function
> around the optimal
> parameters. So if the loss function is quite flat around the
> optimal value, but
> goes up more sharply than the paraboloid found by the Hessian
> matrix around the
> optimal parameters, then the uncertainties could be much larger
> than are really
> I always recommend doing a little Monte Carlo post-mortem to verify
> estimates. Generate parameter values with
> with the optimal parameter values as the mean and cov_beta as the
> matrix. Throw away the values outside of your bounds. Then put the
> parameters into your function and plot all of them against your
> data. If you are
> doing ordinary least squares, it's also quite easy to evaluate the
> function, too, and plot its distribution. It's more difficult to do
> that with
> orthogonal distance regression because the functionality that finds
> orthogonal distances is not exposed by itself.
> Robert Kern
> "I have come to believe that the whole world is an enigma, a
> harmless enigma
> that is made terrible by our own mad attempt to interpret it as
> though it had
> an underlying truth."
> -- Umberto Eco
> Scipy-dev mailing list
More information about the Scipy-dev