Comparing coefficients with different dependent variables
Posted: Fri Jun 10, 2011 10:24 am
I have the following model I use per sector (for example: Industrial, telecom, tech etc.):
cccountry = a + cccountrylag + cccountryfuture + ccworldlag + ccworld + dweekday1-4
cccountry= the daily continuously compounded return on a country sector index
cccountrylag = the daily continuously compounded return on a country sector index with a lag of 1 day
cccountryfuture = the daily continuously compounded return on a country sector index 1 day in the future
ccworld= the daily continuously compounded return on the world index per sector
ccworldlag = the daily continuously compounded return on the world index per sector with a lag of 1 day
dweekday 1-4= dummies for monday, tuesday, wednesday and friday
I run this regression for every sector seperatly, which is obviously the easy part, and then take the residuals of this model and make those the dependent variable in the following model per sector:
e= a + bwin + bloss + bdraw (these are dummies and are the same for every sector)
Run the regressions again per sector and I get nice coefficients per sector. I thought this was sufficient, but my supervisor now says that I need to check the "relative importance" or impact??? of each coefficient compared to the same same coefficient in another sector (so the bloss coefficient of the industrial sector compared to those of the other sectors). My supervisor did not exactly know how to do it himself but he told me that I should find it quite easily on google. I've looking non stop now for 2 days and still have no clue at all. I use panel data for model 1 and then use the residuals I get there in model 2.
If you have any idea, it would be of great help. It was said to be relatively easy once the procedure is known, but like I said before: I am kinda stuck at that part.
Kind regards,
cccountry = a + cccountrylag + cccountryfuture + ccworldlag + ccworld + dweekday1-4
cccountry= the daily continuously compounded return on a country sector index
cccountrylag = the daily continuously compounded return on a country sector index with a lag of 1 day
cccountryfuture = the daily continuously compounded return on a country sector index 1 day in the future
ccworld= the daily continuously compounded return on the world index per sector
ccworldlag = the daily continuously compounded return on the world index per sector with a lag of 1 day
dweekday 1-4= dummies for monday, tuesday, wednesday and friday
I run this regression for every sector seperatly, which is obviously the easy part, and then take the residuals of this model and make those the dependent variable in the following model per sector:
e= a + bwin + bloss + bdraw (these are dummies and are the same for every sector)
Run the regressions again per sector and I get nice coefficients per sector. I thought this was sufficient, but my supervisor now says that I need to check the "relative importance" or impact??? of each coefficient compared to the same same coefficient in another sector (so the bloss coefficient of the industrial sector compared to those of the other sectors). My supervisor did not exactly know how to do it himself but he told me that I should find it quite easily on google. I've looking non stop now for 2 days and still have no clue at all. I use panel data for model 1 and then use the residuals I get there in model 2.
If you have any idea, it would be of great help. It was said to be relatively easy once the procedure is known, but like I said before: I am kinda stuck at that part.
Kind regards,