I am forecasting some price based on VECM. My static forecast works fine, but by adjusting to a dynamic forecast, I get results that are not meaningful.
Here is my code:
// VECM on weekly data for time period 2010-2012 (lag selection based on AIC to avoid serial correlation, rank based on Johansen test)
I now want to forecast lnprice for some time period (2000-2006) prior to the estimation period (2010-2012) to compare to actuals available:
A graph showing the static forecast of lnprice vs. actuals of lnprice shows meaningful results.
Now I want to generate a dynamic forecast by adjusting the last command
A graph showing the dynamic forecast reports some movement in the first periods and then changes to a linear slope.
The -fcast- command does not work as the forecast period is prior to the estimation period.
Any direction how to adjust for a dynamic forecast would be extremely helpful!
Thanks in advance.
Christian
Here is my code:
// VECM on weekly data for time period 2010-2012 (lag selection based on AIC to avoid serial correlation, rank based on Johansen test)
Code:
vec lnprice lnx1 lnx2 lnx3 lnx4 lnx5 if d2010-2012==1, trend(constant) rank(3) lag(4) eststo evecm
Code:
//Create model forecast create fvecm, replace //Add estimation results VECM forecast estimates evecm // 6 endogenous variables added // Identify the five exogenous variables forecast exogenous lnx1 lnx2 lnx3 lnx4 lnx5 // Obtain static forecasts forecast solve, prefix(fcs) begin((w(2000w5))) end((w(2006w52))) static
Now I want to generate a dynamic forecast by adjusting the last command
Code:
forecast solve, prefix(fcd) begin((w(2000w5))) end((w(2006w52)))
The -fcast- command does not work as the forecast period is prior to the estimation period.
Any direction how to adjust for a dynamic forecast would be extremely helpful!
Thanks in advance.
Christian