Option to trace the iterations of dynamic forecasts?
Posted: Thu Dec 03, 2015 1:53 pm
When the FORECAST procedure encounters a missing value during its iterations, it displays output that is very helpful for debugging, e.g.
Is there a way to (optionally) display output like this even when the iteration doesn't encounter a missing value? For example
Although this would create substantially more verbose output, it's extremely helpful for debugging erroneous forecasts in which the forecasted results are incorrect (but not necessarily NA, and therefore not captured by the case of the current debugging output). Is an option like this available?
This is the code I used as an example:
Code: Select all
Forecast for LGDPK solves to NA at 2006:03, on iteration 1
Forecast for LRESINVK solves to NA at 2006:03, on iteration 1
Forecast for GDPK solves to NA at 2006:03, on iteration 1
Code: Select all
Forecast for LGDPK solves to 9.01 at 2006:03, on iteration 3
This is the code I used as an example:
Code: Select all
calendar(q) 1969 1
open data "data.xls"
data(format = xls, org = columns) /
close data
compute begin_samp = 1969:1
compute end_samp = 2005:4
compute end_fcast = 2006:4
smpl begin_samp end_samp
linreg(noprint, define = gdp_eq) lgdpk
# lgdpk{1 to 4} lresinvk{1 to 4} ffedk{1 to 4} constant
linreg(noprint, define = resinv_eq) lresinvk
# lgdpk{1 to 4} lresinvk{1 to 4} ffedk{1 to 4} constant
frml(identity) gdp_level gdpk = exp(lgdpk)
group grp $
gdp_eq>> lgdpk $
resinv_eq>> lresinvk $
gdp_level>> gdpk
forecast(model = grp, from = end_samp+1, to = end_fcast, print)
set ffedk 2006:2 2006:2 = %na
forecast(model = grp, from = end_samp+1, to = end_fcast, print)