[SciPy-User] struggling and fighting with linalg

josef.pktd@gmai... josef.pktd@gmai...
Thu Oct 21 12:02:14 CDT 2010


On Thu, Oct 21, 2010 at 12:46 PM, Warren Weckesser
<warren.weckesser@enthought.com> wrote:
> Hi Josef,
>
> On Thu, Oct 21, 2010 at 11:03 AM, <josef.pktd@gmail.com> wrote:
>>
>> There are a lot of recommendations in the mailing list and recipes in
>> various packages how to use linalg more efficiently.
>> However, yesterday I wasted several hours how to do the multivariate
>> normal distribution with cholesky factor and solve.
>>
>
> What do you mean by "*do* the multivariate normal distribution" (emphasis
> mine)?  numpy.random has multivariate_normal for generating samples; I wrote
> a little cookbook example
> (http://www.scipy.org/Cookbook/CorrelatedRandomSamples) before I realized
> multivariate_normal existed.
>
> But I think you want more than just generating samples, right?

yes, main things I need for statsmodels are loglikelihood and linear
transformation, (and separately in other application non-linear
transformation), plus linear transformation that comes from the
decomposition of the inverse covariance matrix (for whitening or
standardization). The last part is where I got stuck.

The extras will be conditional, marginal distributions, cdf,
confidence regions, ... Part of these just needs to be collected (and
tested).

Josef

>
> Warren
>
>
>
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>


More information about the SciPy-User mailing list