Urgent !!! Coefficient Restriction

[londonhaha]

Hi there !

I am currently using EVIEWS 8.

I have a multiple regression model estimated by OLS below:

DLTHB = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP + C(6)*RESID_CNY

DL is log difference. I would like to put restriction on coefficients that C(2)+C(3)+C(4)+C(5)+C(6) = 1

What full code should I put in the EVIEWS estimation command box?

Thanks in advance.

[EViews Gareth]

DLTHB = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP + (1-c(1)-c(2)-c(3)-c(4)-c(5))*RESID_CNY

[londonhaha]

Thanks for the answer.

But what is the code to put in estimation command box?

For example, I need to write LS DLTHB C DLUSD DLEUR DLJPY DLGBP RESID_CNY to estimate OLS regression.

[EViews Gareth]

The code I gave

[londonhaha]

Thanks again!

I mean the sum of the coefficients = 1 and this shouldn't include constant term C(1).

The code you gave is:

DLTHB = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP
+ (1-C(1)-C(2)-C(3)-C(4)-C(5))*RESID_CNY

I want to have only sum of coefficients ( DLUSD, DLEUR, DLJPY, DLGBP and Resid_CNY) not constant.

Should I include C(1) into the coefficient of Resid_CNY---> (1-C(1)-C(2)-C(3)-C(4)-C(5)) ???

Sorry I am confused.

[EViews Gareth]

Sorry, I missed that. Yeah, don't include c(1)

[londonhaha]

So should it be :

DLTHB = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP
+ (1-C(2)-C(3)-C(4)-C(5))*RESID_CNY

?

[londonhaha]

I tried this following code:

DLTHB = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP
+ (1-C(2)-C(3)-C(4)-C(5))*RESID_CNY

And the result is:

Dependent Variable: DLTWD
Method: Least Squares
Date: 11/18/15 Time: 13:00
Sample (adjusted): 1/03/1997 11/16/2015
Included observations: 4730 after adjustments
HAC standard errors & covariance (Bartlett kernel, Newey-West fixed
bandwidth = 10.0000)

DLTWD = C(1) + C(2)*DLUSD + C(3)*DLEUR + C(4)*DLJPY + C(5)*DLGBP
+ (1-C(2)-C(3)-C(4)-C(5))*RESID_CNY

Coefficient Std. Error t-Statistic Prob.
C(1) 3.44E-05 4.23E-05 0.812020 0.4168
C(2) 0.824438 0.012849 64.16204 0.0000
C(3) 0.010291 0.007749 1.328057 0.1842
C(4) 0.039601 0.008875 4.461927 0.0000
C(5) 0.047630 0.010787 4.415690 0.0000


R-squared 0.877613 Mean dependent var 1.92E-05
Adjusted R-squared 0.877510 S.D. dependent var 0.008188
S.E. of regression 0.002866 Akaike info criterion -8.871034
Sum squared resid 0.038800 Schwarz criterion -8.864204
Log likelihood 20985.00 Hannan-Quinn criter. -8.868633
F-statistic 8470.529 Durbin-Watson stat 2.048033
Prob(F-statistic) 0.000000 Wald F-statistic 4582.838
Prob(Wald F-statistic) 0.000000



You can see that there is no variable Resid_CNY in the regression result. Is this correct?

[EViews Gareth]

Yes