Page 1 of 1

Different GARCH results when sorted by ascending/descending

Posted: Sun Aug 12, 2012 2:54 am
by chauhan7
So I'm currently using an out of date Eviews 6, although I have tested this on an up to date Eviews 7 at university, and get the same effect.

I have attached a sample which shows two sets of 300 stock observations with their arithmetic returns and dummy variables to signify each day of the week, where both sets of 300 are sorted in a different way. When doing the GARCH model based on "return monday tuesday wednesday thursday friday" I get different results for both sets, when i should not?

Any help?

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:15 am
by EViews Gareth
Could you provide the EViews file, plus instructions on what you're doing. It is hard to make out what you've done from your description/data.

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:27 am
by chauhan7
I have uploaded both sets of data in workfile form. Also what I am doing is looking at the returns for the FTSE 100 and examining the daily market anomalies (i.e. Friday effect where returns on friday are significantly higher than those of the other days). However, when I sorted the data in excel the other way around (whilst adjusting the returns formula appropriately) the output was different. I hope that makes sense.

By output i mean the probability values of the z-scores, ARCH test, and squared residual values.

(note: the original sample is much larger than this, I just took a sample to make it simple

Edit: reuploaded test 2 with GARCH

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:31 am
by EViews Gareth
What are you doing in EViews?

Neither of those workfiles are structured by date? Is there a reason why?


Just as an FYI, I put your two sets of data from Excel into EViews (just by copying the entire block and pasting as a new workfile) and estimated the specification you gave above as GARCH, and got identical results in each.

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:32 am
by chauhan7
What are you doing in EViews?

Neither of those workfiles are structured by date? Is there a reason why?


Just as an FYI, I put your two sets of data from Excel into EViews (just by copying the entire block and pasting as a new workfile) and estimated the specification you gave above as GARCH, and got identical results in each.
because the stock market dates contain some missing days such as weekends, holidays and other days where the data was unavailable, so I didnt think I could do that?

Different GARCH results when sorted by ascending/descending

Posted: Sun Aug 12, 2012 3:33 am
by EViews Gareth
Sure you can.

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:36 am
by chauhan7
What are you doing in EViews?

Neither of those workfiles are structured by date? Is there a reason why?


Just as an FYI, I put your two sets of data from Excel into EViews (just by copying the entire block and pasting as a new workfile) and estimated the specification you gave above as GARCH, and got identical results in each.
So did you create an unstructured workfile, open an empty group and simly copy the data into there? Did you see my GARCH outputs for both files?

Different GARCH results when sorted by ascending/descending

Posted: Sun Aug 12, 2012 3:38 am
by EViews Gareth
Copy it in Excel, then right click on an empty part of the EViews desktop and select Paste as New Workfile

Re: Different GARCH results when sorted by ascending/descend

Posted: Sun Aug 12, 2012 3:46 am
by chauhan7
Hmm, so I tried what you did on my university computer and it worked! It picked up the dates from excel, and the output was the same and it picked up the date values.

Will test out my full sample data now.

Thanks!