please help me out..........

[jiya]

i estmated the followng model with the objective to find compound growth rate of y:
ln(y)= a+bt+cD+dtD
where, D=dummy,
t=time.
Now, compound growth rate(CGR)= antilog(c+d)-1*100............forD=1
compound growth rate=antilog(c)-1*100................for D=0
But the problem is that coefficient of tD i,e; d found to be statistically insignificant event at 50% level of significance. Should i calculate CGR by antilog(c+0)-1*100............forD=1?..............or i should still use antilog(c+d)-1*100 to find CGR and mention in the footnote that coefficient of tD is statistically insignificant.

[trubador]

If it is statistically insignificant with such a low level of confidence, then it means that the variable of your concern does not have any important explanatory power and you should drop it from the equation altogether. However, if it has to have an important meaning from the theoratical point of view, then you should overhaul your equation (e.g. add/drop variables, check for multicollinearity, etc.) and specify another one. In your case, as I see it, there is simply no interaction between dummy and time variables, so I would continue with the parsimonious specification.

[jiya]

tx a lot

[jiya]

hi............i want to estimate a function viz one-period % change (in decimal) using Eviews4 and for that i used eviews specification y c pch(x).......but it says "pch" is not defind. Can anybody plz tell me the correct specification...................

[trubador]

I think the correct use would be @pch(x).

[jiya]

tx a lot.

[jiya]

hi.........m estimating a model , lnY=C+Bt, where Y stands for GDP and t stands for time. Where i found that coeff of t is highly sig. but my result suffers from autocorrelation sinnce the value of D/W is only 0.32. So i opted to correct that by re-specifying the model as lnY c t ar(1) .After that correction the coeff. of t becomes insig and in the bottom of output window (Eviews4.1) it is written "Estimated AR process is nonstationary". Will u plz let me kw the meaning of this statement.

[tchaithonov]

Your model is not I(0) .. that means you should be looking for a ARI(p,d) solution .. do a correlogram and look at the ACF. Then, you need to proceed with a dlog(Y, d)... (d = # of integrated period)..

[jiya]

i want to test Granger Causality using Eviews4 between two variables say X and Y both the series being I(1) one at lag 2 and other at lag 4. Will any body let me know hw should i chose lags in testing Granger Causality given the fact that one can not see the value of AIC and SIC in the dialog box that appears in Pairwise Granger Causality Test in eviews?