bandwidth and the number of lags

For requesting general information about EViews, sharing your own tips and tricks, and information on EViews training or guides.

Moderators: EViews Gareth, EViews Moderator

RDS
Posts: 86
Joined: Thu Feb 17, 2011 9:50 am

bandwidth and the number of lags

Postby RDS » Fri Nov 08, 2019 3:41 am

I would like to know how I can infer the number of lags used for the bandwidth associated to the (long-run) variance estimator.

In running a system with GMM, I get the following output:

Included observations: 247
Total system (balanced) observations 3458
Kernel: Bartlett, Bandwidth: Variable Newey-West (2), No prewhitening
Simultaneous weighting matrix & coefficient iteration
Convergence achieved after: 1 weight matrix, 2 total coef iterations

Can you tell me if I can infer the number of lags from the Bandwidth (Variable Newey-West (2)).
Is it 2 the number of lags?

For your additional information, if I use the NW fixed option with 247 observations, I get the following, where the Bandwidth is 5.
Using the formula available here is rather complex to extract the number of lags: http://www.eviews.com/help/helpintro.ht ... 23ww155429

Included observations: 247
Total system (balanced) observations 3458
Kernel: Bartlett, Bandwidth: Fixed (5), No prewhitening
Simultaneous weighting matrix & coefficient iteration
Convergence achieved after: 1 weight matrix, 2 total coef iterations

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13305
Joined: Tue Sep 16, 2008 5:38 pm

Re: bandwidth and the number of lags

Postby EViews Gareth » Fri Nov 08, 2019 7:58 am

Yes, 2 is the number of selected lags.
Follow us on Twitter @IHSEViews

RDS
Posts: 86
Joined: Thu Feb 17, 2011 9:50 am

Re: bandwidth and the number of lags

Postby RDS » Sat Nov 09, 2019 4:47 am

If I run an OLS regression, I get the following:

Included observations: 247
HAC standard errors & covariance (Bartlett kernel, Newey-West automatic bandwidth = 7.7735, NW automatic lag length = 4)
No d.f. adjustment for standard errors & covariance

If I apply a rule of thumb from the literature I get a Bandwidth = 4.7-4.9.

How does all this square? Thank you

RDS
Posts: 86
Joined: Thu Feb 17, 2011 9:50 am

Re: bandwidth and the number of lags

Postby RDS » Tue Nov 12, 2019 3:41 pm

Possibly I was not clear.

If I run an OLS regression with 247 observations, I get the following

Newey-West automatic bandwidth = 7.7735
NW automatic lag length = 4

What is the number of lags that I can infer from the Bandwidth?
In the OLS case the number of lags seems to be associated with a bandwidth = 7.7735.

While when running a GMM, as you said, bandwidth = number of lags = 2.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13305
Joined: Tue Sep 16, 2008 5:38 pm

Re: bandwidth and the number of lags

Postby EViews Gareth » Tue Nov 12, 2019 4:33 pm

Sorry, I didn't notice you had a system. 5 is the bandwidth.
Follow us on Twitter @IHSEViews

RDS
Posts: 86
Joined: Thu Feb 17, 2011 9:50 am

Re: bandwidth and the number of lags

Postby RDS » Thu Mar 26, 2020 7:44 am

If I run an OLS with 247 observations with a Bandwidth method being Newey-West automatic, I get the following:

HAC standard errors & covariance (Bartlett kernel, Newey-West
automatic bandwidth = 9.8793, NW automatic lag length = 4)


If I run an OLS with 247 observations with a Bandwidth method being Newey-West fixed, I get the following:

HAC standard errors & covariance (Bartlett kernel, Newey-West fixed
bandwidth = 5.0000)


Can you tell me in the latter case what is the NW automatic lag length?


Return to “General Information and Tips and Tricks”

Who is online

Users browsing this forum: No registered users and 7 guests