pacf has incorrect standard errors for lag 0 #1969

Closed
bashtage opened this Issue Sep 19, 2014 · 5 comments

Projects

None yet

2 participants

@bashtage
Contributor

Lag 0 standard errors should be both 1, like how it appears in acf

@josef-pkt
Member

Do you mean confidence intervals ?
I don't see standard errors returned anywhere directly.

>>> sm.tsa.pacf(dta, alpha=0.05)[1][:5]
array([[ 0.88850154,  1.11149846],
       [ 0.71136583,  0.93436274],
       [-0.80180867, -0.57881176],
       [-0.25207058, -0.02907367],
       [-0.05541805,  0.16757886]])
>>> sm.tsa.acf(dta, alpha=0.05)[1][:5]
array([[ 1.        ,  1.        ],
       [ 0.70870284,  0.93169975],
       [ 0.28050977,  0.62202721],
       [-0.14541504,  0.22456814],
       [-0.46088878, -0.09069514]])
@bashtage
Contributor

Yes, CI is returned but the standard error that goes into the CI is incorrect at lag 0 for pacf. The 0 lag CI should always be [1,1], like in acf since the standard error is always 0 at this lag.

@josef-pkt josef-pkt added this to the 0.6 milestone Sep 20, 2014
@josef-pkt
Member

@bashtage Do you want to prepare a PR?
Otherwise, I add it to a misc bugfix PR that I'm planning for today.

@bashtage
Contributor

Will leave this one to you if it is easy.

This was referenced Sep 22, 2014
@josef-pkt
Member

fixed in da0fe2e PR #1987
typo in the commit message that referred to unrelated PR

@josef-pkt josef-pkt closed this Sep 23, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment