[Numpy-discussion] partial autocorrelation function

[Skipper Seabold]

On Thu, Mar 25, 2010 at 11:25 AM,  <josef.pktd at gmail.com> wrote:
> On Thu, Mar 25, 2010 at 11:20 AM, Skipper Seabold <jsseabold at gmail.com> wrote:
>> On Thu, Mar 25, 2010 at 11:10 AM, Mark Bakker <markbak at gmail.com> wrote:
>>> Anybody know of a partial autocorrelation function in numpy? Maybe in scipy?
>>>
>>> Thanks,
>>>
>>> Mark
>>>
>>> ps. I know, not too difficult, but if it is around I'd be happy to use it.
>>>
>>
>> I'm sure there's another version somewhere, but I couldn't find one
>> and wrote pacorr
>>
>> http://bazaar.launchpad.net/~jsseabold/statsmodels/statsmodels-skipper/annotate/head%3A/scikits/statsmodels/sandbox/tsa/stattools.py#L92
>>
>> It relies on statsmodels, but could easily be made to work without it.
>>  You can roll your own OLS,  add_constant just adds a column of ones,
>> and you can see lagmat here:
>>
>> http://bazaar.launchpad.net/~jsseabold/statsmodels/statsmodels-skipper/annotate/head%3A/scikits/statsmodels/sandbox/tools/tools_tsa.py
>
> My versions are in the last release
>
> scikits.statsmodels.sandbox.tsa.pacf_ols
> scikits.statsmodels.sandbox.tsa.pacf_yw
>
> difference in the third (I think) decimal compared to matlab, where I
> didn't find out the reason
>
> Josef
>

Ah, there they are.  To be clear, use these then, as I'll be removing the other.

Skipper