Parameter estimation for ODEs- negative values on covariance diagonal, high values of std error

[Garima Singh]

Hi,

I am trying to fit a set of two odes using orient from SciPy and lmfit minimise. I get a nice fit, but too high std errors and in many cases, a nan values. This is because of negative value on the diagonal of the covariance matrix.

I would appreciate insights/suggestions in the regard. 

[Renee Otten]

hi Garima,

us usual, providing a minimal, working example would possibly help to provide a useful answer. From your current question the only thing one can probably say right now is something like “avoid nan values in your model function”. 

On Jul 7, 2021, at 8:27 AM, Garima Singh <garima....@gmail.com> wrote:

Hi,

I am trying to fit a set of two odes using orient from SciPy and lmfit minimise. I get a nice fit, but too high std errors and in many cases, a nan values. This is because of negative value on the diagonal of the covariance matrix.

I would appreciate insights/suggestions in the regard. 

--
You received this message because you are subscribed to the Google Groups "lmfit-py" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lmfit-py+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lmfit-py/8fb06468-a9b6-489f-ae26-92b98ab59d6fn%40googlegroups.com.

[Matt Newville]

On Wed, Jul 7, 2021 at 7:27 AM Garima Singh <garima....@gmail.com> wrote:

Hi,

I am trying to fit a set of two odes using orient from SciPy and lmfit minimise. I get a nice fit, but too high std errors and in many cases, a nan values. This is because of negative value on the diagonal of the covariance matrix.

How do you get negative values on the diagonal of the covariance matrix?  I don't know how one could interpret variances that are negative.

--Matt