EViews.com

Hi, i have the following problem....would be great if someone can help

My plan is to test if my international benchmakr is efficient.

So i take the constituents of my benchmark, in this case time series for 23 countries and run each country against my international benchmark (msci_world):

equation eq_{%name}.ls(n) {%name} c msci_world

So i test for every country if my country-alpha is signifikant (23 individual regressions for 23 countries).

Here my code:
for !i = 1 to country.@count
%name = country.@seriesname(!i)

equation eq_{%name}.ls(n) {%name} c msci_world

!s= eq_{%name}.@regobs - eq_{%name}.@ncoef
table countrydata
countrydata(!i,1) = eq_{%name}.@coef(1)
countrydata(!i,2) = eq_{%name}.@coef(2)
countrydata(!i,3) = eq_{%name}.@tstat(1)
countrydata(!i,4) = eq_{%name}.@tstat(2)
countrydata(!i,5) = @tdist(eq_{%name}.@tstat(1),!s)
countrydata(!i,6) = @tdist(eq_{%name}.@tstat(2),!s)
countrydata(!i,7) = eq_{%name}.@r2
countrydata(!i,8) = eq_{%name}.@rbar2

next

Question:

How can i test if all the alphas together are significant. So i don't just want to receive the country individual alpha, but also an alpha and a test for the whole set of 23 equtions (equation eq_{%name}.ls(n) {%name} c msci_world). I think i can use the wald test, but how do i get the test to include all equations together?

Is there a way to include it in my written code?

Thanks for your help!

Set up a system and run a seemingly unrelated regression.

Thanks for the quick response.

Could you please tell me one further thing:

Why are my results (alpha, p-values) from the regression equation eq_{%name}.ls(n) {%name} c msci_world different from the results i get by executing the regression in a system ("seemingly unreleated regression").

is it because i use the (n) in my equation eq_{%name}.ls(n) (HAC)? How can i adjust my system to get the same results in my system as i get from the original regression?

Thanks for your help!

Using HAC would affect the standard errors, but not the coefficients. Since you've set up a system, you can also estimate that system by OLS if you want. Generally though, SUR is more efficient.

one more question regarding wald test....to get help would be really great!

When i set up the system and run an seemingly unrelated regression, does it matter to do it with 25 or 300 time series in terms of the quality of my results?

Additionally i run the regression für 300 investment funds. Because i change my benchmark (add another coefficient or two) i want to test, if the additional coefficient(s) contributes significantly to my results. So i do a wald test. To do it individually for every investment fund (by choosing wald test in the menu) is very tedious. Is there a way to integrate the wald test in my program by writing a command line (maybe there is a abbeviation for the wald test)? Could you please help me out with the command line for my program?

I saw that there is another coefficient test called redundant variable - likelihood ratio (in the menu coefficient test). What is the difference to the wald test? When i want to test if additional coefficients help to exlpain the return of investment funds should i use the former or the latter test?

Thanks for your help!