EViews.com

Hi,
I''m new with eviews and I should estimate an exponential trend using non linear least squares in order to compare this kind of model with a linear and a quadratic model. The model should be lny=a+btrend+error and I tried to estimate an equation like this one (after generating a series t=@trend) y=c(1)+exp(c(2)*@trend). At the beginning it worked (I have to do the estimate for three different variable) but with the other two variables I ran into "overflow message'" or weird estimates. For example there were stats like these:


Dependent Variable: GDP_PER_CAP
Method: Least Squares
Date: 09/11/12 Time: 11:33
Sample: 1960 2009
Included observations: 50
Convergence achieved after 1 iteration
GDP_PER_CAP=C(1)+EXP(C(2)*@TREND)

Coefficient Std. Error t-Statistic Prob.

C(1) 149.3340 6.9E+110 2.2E-109 1.0000
C(2) 5.287790 0.002976 1776.886 0.0000

R-squared ####### Mean dependent var 16358.06
Adjusted R-squared ####### S.D. dependent var 13854.51
S.E. of regression 4.9E+111 Akaike info criterion 517.2093
Sum squared resid 1.1E+225 Schwarz criterion 517.2858
Log likelihood -12928.23 Hannan-Quinn criter. 517.2384
Durbin-Watson stat 0.989920

I don' t really know what is the problem, could you please help me?

One possibility is that you need to set the C vector to a reasonable value before the estimate. And do you want a coefficient in front of exp?

Yes I do. How can I set the C vector to a reasonable value?

Use the values from your log equation.