Confidence interval for PACF

[martin.g25]

Hello, could anybody please explain me how the confidence interval (the blue background on the plot) is calculated in the PACF plot? In my case it should be 0.3162 and -0.3162.

Thank you.

 

[Harry_the_cat]

What % CI?

 

[BigBeachBanana]

Hello, could anybody please explain me how the confidence interval (the blue background on the plot) is calculated in the PACF plot? In my case it should be 0.3162 and -0.3162.

Thank you.

I'm not so sure that's a CI. The PACF is a descriptive statistic that summarizes the properties of a given time series, and it is not a population parameter that can be estimated from a sample of data. Therefore, it is also not appropriate to use a confidence interval to estimate the uncertainty in the value of the PACF.

 

[martin.g25]

Maybe it is not a confidence level. I am not sure, therefore I am asking...

For the blue background there is a pertaining label "Two Standard Errors". Do this help?


I get the plots using this simple SAS code:

proc arima data=VZORKA plots(unpack)=all;
identify var="YOUR_VARIABLE" nlag=12;
ods output SeriesACFPlot=SAP SeriesPACFPlot=SPP;
run;

 

[BigBeachBanana]

Maybe it is not a confidence level. I am not sure, therefore I am asking...

For the blue background there is a pertaining label "Two Standard Errors". Do this help?


I get the plots using this simple SAS code:

proc arima data=VZORKA plots(unpack)=all;
identify var="YOUR_VARIABLE" nlag=12;
ods output SeriesACFPlot=SAP SeriesPACFPlot=SPP;
run;

I've looked into it a bit more. Though it's uncommon, yes, you can calculate the confidence interval for PACF but at each lag

For lag [imath]k[/imath] and sample size [imath]N:[/imath]
[math]SE(k) = \sqrt{\dfrac{1}{N}}[/math][math]CI = PACF(k) \pm z_{\alpha/2} \times SE(k)[/math]
I'm unfamiliar with SAS, so can't comment on the "standard errors"

 

[martin.g25]

Based on your formula I would get different CL level for different PACF values. However, on the plot you can see that the "blue background" has the same boundaries for each PACF value. These boundaries are 0.3162 and -0.3162.

Or am I missing something?

Thank you.

 

[BigBeachBanana]

Based on your formula I would get different CL level for different PACF values. However, on the plot you can see that the "blue background" has the same boundaries for each PACF value. These boundaries are 0.3162 and -0.3162.

Or am I missing something?

Thank you.

That's the point. The CI varies for each lag. So whatever that box is, it's not CI.