EViews.com

Hi,

I'm an absolute beginner and using EViews 7 at the moment.
I'm running a cross-sectional regression with the dependent variable is returns of sector indices and the regressors are sector and country dummies (=1 if that sector index belongs to country x, 0 otherwise. similarly to sector variables).
My data is as follow in Excel (e.g. 3 countries, 10 sector indices in each country, time-series vary for different countries).



(Please note: some time-series data is unavailable for certain countries so I have an unbalanced panel).
I have the following in EViews:



Could you please show me how to:

1. Create sector/country dummies given my current format of data.
2. How can I specify my regression to be Weighted Least Square? And how can I put 2 restrictions that the sum of weights of all country coefficents = 0, similarly with sector coefficients?
3. Will the fact that some of the time series data is missing (e.g. some countries without HC (Healthcare) sector indices) affect my regression? And will EViews treat those missing as 'no data' or 'zero'?
Could you please advise on how to fix this problem?

I would be very thankful for your help and I look forward to hearing from you.

Somehow the images don't work.
Could you please follow these links instead?

http://www.flickr.com/photos/77043061@N ... otostream/
http://www.flickr.com/photos/77043061@N ... otostream/

1. You'll have to provide a little more detail. You have a panel data set (cross-sections and time). What do you mean when you say you are running a cross-sectional regression?

2. EViews doesn't support weighted least squares in panels.

3. It will not use an observation for which any of the variables has an NA.

1. You'll have to provide a little more detail. You have a panel data set (cross-sections and time). What do you mean when you say you are running a cross-sectional regression?

2. EViews doesn't support weighted least squares in panels.

3. It will not use an observation for which any of the variables has an NA.

Hi Gareth,

Thank you for the reply. The regression that I'm running is:

Ri = alpha + Beta1*I1 + Beta2*I2 + ... + Gamma1*C1+ Gamma2*C2 + ... + errors

where: I1,I2... = industry dummies (=1 if Ri belongs to that industry, 0 otherwise)
C1, C2,... = country dummies

How can I set up these dummies?

As 1 return always belongs to 1 country & 1 industry (perfect multicollinearity), I want to set up 2 restrictions:
The sum of Wi*Beta-i = 0 , sum of Wj*Gamma-j = 0 (where Wi, Wj = market-cap weights of industries, countries). How can I do this assuming that I have data of those weights.

Final question: will the fact that my panel data is unbalanced (some countries with data from 1995-2012, some from 2000-2012) affect EViews in running the regression?

Many thanks,

Nancy

I still don't see where time comes into your regression.

The time-series is mainly for return calculations I suppose.
As I want to regress returns of each industry index against a set of country and industry dummies.
My main issue here is how to create these dummies (as described above).

I'm just trying to replicate an academic paper and my apologies for making it confusing.

Cheers,

Nancy

Take a look at the @expand function.