serial correlation & heteroskedasticity

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 6:11 am

Dear all,

I hope that somebody can help me. My regression model is serially correlated. However, I cannot correct it (have tried to run it with lagged variables and also with AR(1) and with the Newest-West button)

what can I further do? Moreover, the model is also heteroskedastic..
Please help me!

Regards

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 6:38 am

How will be serial correlation and heteroskedasticity corrected with the NEWEY WEST button? I'm using Eviews 6 and if I activate the button - and run again the LM Serial Correlation test . the results are the same (serially correlated)..

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 7:03 am

Newey-West corrects standard errors for serial correlation and heteroskedasticity. It doesn't change the coefficients or eliminate the serial correlation.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 7:43 am

thank you for your answer.

so if my regression is serially correlated and heteroskedastic - is it enough to active the NEWEY WEST button? Will this solve the problem of serial correlation and heteroskedasticity? Moreover, can I use the results (residuals of this regression model) for e.g. an error correction model? ?

Thank you very much!

Regards

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 7:58 am

I suppose it depends what the "problem" is that you need to solve.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 9:31 am

My regression model is:
lnp c lncr lnex where p=price cr=crude oil and ex= exchange rate

However, this model is serially correlated and heteroskedastic.. I've tried to solve the serial correlation and heteroskedasticity problem but it wasn't possible...
I have to run an error correction model with the residuals of this OLS model, but I couldn't because of these problems above. So my question is if it's enough to estimate with HAC Newey West (activating the button)? I don't know how to solve it:/

Thank you very much!

Regards

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 9:45 am

1ci wrote:My regression model is:
lnp c lncr lnex where p=price cr=crude oil and ex= exchange rate

However, this model is serially correlated and heteroskedastic.. I've tried to solve the serial correlation and heteroskedasticity problem but it wasn't possible...
I have to run an error correction model with the residuals of this OLS model, but I couldn't because of these problems above. So my question is if it's enough to estimate with HAC Newey West (activating the button)? I don't know how to solve it:/

Thank you very much!

Regards

Using HAC Newey-West has no effect whatsoever on the residuals.
I don't know whether residuals being serially correlated and heteroskedastic is a problem for whatever you're trying to do, but if it is then nothing is fixed by using Newey-West.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 9:57 am

thank you for your answer.

Can you probably tell me how to solve serial correlation and heteroskedasticity problem? I've daily data and using lagged variables or AR(1) couldn't solve the problem..
It would be great if you can help me...
Thank you very much.

Regards

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 9:59 am

You might post your workfile to see if anyone has a suggestion. Be sure to include in the workfile the equation object that you estimate.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 10:12 am

Thank you for your answer.

Attached is the excel sheet with the data where p=price, cr=crude oil and ex=exchange rate
daily data from 01. September 2011 to 02. April 2013

my regression equation is:

log(p) c log(cr) log(ex)

however, I couldn't solve the serial correlation and heteroskedasticity problem..
Hope, you can help me.

Thank you very much.

Regards
Attachments
Data.xlsx
(71.86 KiB) Downloaded 437 times

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 10:14 am

It would be easier if you posted your EViews workfile, and included in the file the equation you have estimated.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 10:20 am

Sorry ..

thank you very much!
Attachments
analysis juni.wf1
(103.23 KiB) Downloaded 400 times

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 10:23 am

unfortunately, I haven't got the Eviews Program now..

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: serial correlation & heteroskedasticity

Postby startz » Mon Jun 10, 2013 10:30 am

Your variables probably have unit roots. This may account for the serial correlation. I don't see a way for you to eliminate the serial correlation.

1ci
Posts: 24
Joined: Tue May 07, 2013 7:56 am

Re: serial correlation & heteroskedasticity

Postby 1ci » Mon Jun 10, 2013 10:34 am

yes my variables are I(1)
can I still go on with the error correction model?


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 27 guests