PANEL DATA

[ec07ssn]

hi
i have 6 years observations for 59 stocks in UK. and each year has a weekly observations.
my format in excel looks something like this:
COL A COL B COL C COL D COL E
date s tock Ri Rf Rm - Rf
06/01/03
13/01/03
.
.
.
22/12/09
29/12/09

how do i structure this as a panel data in eview6 and run ONE regression on my data as a whole. thanks

[tchaithonov]

I think you can just insert the data into Eviews and change the structure of the data to balanced panel with date as date id and code as cross section id. Just make sure that you don't have any space in between your column names... your Rm-Rf may not work as variable name because of the minus in the middle.

[EViews Gareth]

http://forums.eviews.com/viewtopic.php?f=7&t=74

[ec07ssn]

Thanks!
But when i insert data into eviews and try to restructure then theres no option for balanced data
so what i try to do is create a new balanced panel workfile
you know when it asks by observation or by series, which one should that be?
what should be the name for series or number if named in file.
sorry im really really new to eviews. i would really appreciate if u can help me do it step by step. thanks

[ec07ssn]

and how come the dates in my data doesnt match with the dates in the eviews.
and why do i get NAs instead of the observations

[EViews Gareth]

You're going to have to be more precise on exactly what you're doing.

When I open up that Excel file in EViews 7, EViews correctly interprets it as a Panel without requiring any further work on my behalf.

[ec07ssn]

i am using eviews 6 and my equation is Ri = Rf + B(Rm-Rf)
i have to regress this equation in eviews using the data i have and estimate B.. and because i have both cross sectional and time series i have absolutely no clue what to do.
can you please tell me how to do this as im really new to this!

[ec07ssn]

hi can u please tell me how to this, in detail! thanks

[EViews Gareth]

I think you're going to have to spend some time writing a coherent question. At the moment it really isn't very clear exactly what you have done, what you are trying to do, and why what you have done is not working for you. If you are able to explain your predicament in detail, perhaps people will be able to respond to you in detail.

[tchaithonov]

I think I just spotted the answer to the question -- the excel file is for Excel 2007 and Eviews 6 doesn't support 2007!

Convert your file to *.xls before you import it into Eviews. That should solve your problem.

[EViews Gareth]

Nice catch.

[ec07ssn]

i have already converted the file into xls.
But the prob is..
i dont know how to sturcture it propery and estimate the panel regression.
when i jst enter this data into eviews6 and without restructuring i run a regression, the t stat is toooooo high..
how do i restructure it..
from the data i have column c as my y dependent variable and column d and column e as my explanatory variable.
But because my stock is getting repeated and i have more than one observation for each stock as this is weekly 6 years data, each stock gets repeated...
so how do i restructure it properly and estimate an equation in such a way that i dont get drastic results.
Thanks

[tchaithonov]

There are 10 weeks in your dataset that do not start on a Monday:

1/6/2003
5/6/2008
5/27/2008
6/10/2008
8/26/2008
4/14/2009
5/5/2009
5/26/2009
9/1/2009
12/29/2009

These are all Tuesdays. That's why Eviews can't tell whether you want it to be balanced or unbalanced. Next time, check your dates first before you even open eviews.

[ec07ssn]

oh...
so what do i do now.

[tchaithonov]

Here is an excel file with the modified date (with the formulas in it to convert dates), and the resulting eviews file. The t-stat is too high and you should be able to google for an answer like this:

"The variance reported by econometric packages includes the sampling variance but not that part of the variance due to the common error. Thus if there are any shocks that are correlated within year/group cells, the reported t-statistic will be too high. We will tend to find an effect of the treatment even if
none exists."

Donald & Lang - Inference with Difference in Differences and Other Panel Data
http://128.197.153.21/lang/MS10131.pdf

I am not sure if this is the real reason but that gives you something to start with. Gareth, would you please confirm if this is the reason for the t-stat issue? Thanks.