Dear stata listers,
I'm currently doing an event study using GARCH model to test the effects of certain events on the volatility of stock markets.
After the GARCH model I need to predict the variance. However, my problem is that I need a dynamic prediction and not a one-day-ahead prediction.
The estimation window is t=-500 to t=-1 and the event window: t=0 to t=30. I want the dynamic prediction to start at the day of the event (t=0) using the information set available on t=-1 (the last day of the estimation window)
I'm using stata 15
My statacodes look like:
gen predicted_return=.
gen predicted_variance=.
gen nIndicator = _n
tsset nIndicator
egen id=group(company_id)
forvalues i=2(1)2 { /*note: replace N with the highest value of id */
l id company_id if id==`i' & dif==0
arch ret market_return if id==`i' & estimation_window==1, arch(1) garch(1) nolog
predict p if id==`i'
replace predicted_return = p if id==`i' & event_window==1
drop p
local tInd = nIndicator[dif==0 & id==`i']
predict p if id==`i', variance dynamic(`tInd')
replace predicted_variance = p if id==`i' & event_window==1
drop p
}
However, when I do this the dynamic prediction starts at t-500 whilst it needs to start at t (the day of the event).
and my data is as follows:
date market_return country_id ret set event_date company_id datenum td dif event_window count_event_obs estimation_window count_est_obs predicted_return predicted_variance nIndicator id
26aug2015 2.06% 2 -.0232401 37 02sep2015 84 778 783 -5 0 30 1 500 9892 7
27aug2015 2.11% 2 .0029889 37 02sep2015 84 779 783 -4 0 30 1 500 9893 7
28aug2015 0.46% 2 .0216223 37 02sep2015 84 780 783 -3 0 30 1 500 9894 7
31aug2015 -0.76% 2 .0295807 37 02sep2015 84 781 783 -2 0 30 1 500 9895 7
01sep2015 -2.72% 2 .0182079 37 02sep2015 84 782 783 -1 0 30 1 500 9896 7
02sep2015 0.93% 2 -.0117279 37 02sep2015 84 783 783 0 1 30 0 500 9897 7
03sep2015 0.45% 2 -.0159718 37 02sep2015 84 784 783 1 1 30 0 500 9898 7
04sep2015 -1.68% 2 .003991 37 02sep2015 84 785 783 2 1 30 0 500 9899 7
07sep2015 0.18% 2 .0183392 37 02sep2015 84 786 783 3 1 30 0 500 9900 7
02dec2014 0.35% 2 -.0107188 50 10dec2014 97 597 603 -6 0 30 1 500 11230 8
03dec2014 0.22% 2 -.0087944 50 10dec2014 97 598 603 -5 0 30 1 500 11231 8
04dec2014 -0.18% 2 -.0024023 50 10dec2014 97 599 603 -4 0 30 1 500 11232 8
05dec2014 0.11% 2 .0015828 50 10dec2014 97 600 603 -3 0 30 1 500 11233 8
08dec2014 -0.69% 2 -.0045892 50 10dec2014 97 601 603 -2 0 30 1 500 11234 8
09dec2014 -0.26% 2 -.0043328 50 10dec2014 97 602 603 -1 0 30 1 500 11235 8
10dec2014 -1.37% 2 -.0221322 50 10dec2014 97 603 603 0 1 30 0 500 11236 8
11dec2014 0.04% 2 -.0354491 50 10dec2014 97 604 603 1 1 30 0 500 11237 8
12dec2014 -1.41% 2 -.0096146 50 10dec2014 97 605 603 2 1 30 0 500 11238 8
15dec2014 -1.22% 2 -.0088118 50 10dec2014 97 606 603 3 1 30 0 500 11239 8
16dec2014 0.06% 2 -.0230455 50 10dec2014 97 607 603 4 1 30 0 500 11240 8
17dec2014 1.02% 2 -.0277923 50 10dec2014 97 608 603 5 1 30 0 500 11241 8
30nov2015 -0.35% 2 -.0083918 1587 03dec2015 1634 843 846 -3 0 30 1 500 711735 469
01dec2015 0.93% 2 -.0026178 1587 03dec2015 1634 844 846 -2 0 30 1 500 711736 469
02dec2015 -0.81% 2 -.0009559 1587 03dec2015 1634 845 846 -1 0 30 1 500 711737 469
03dec2015 -1.06% 2 .0099418 1587 03dec2015 1634 846 846 0 1 30 0 500 711738 469
04dec2015 0.99% 2 .0026078 1587 03dec2015 1634 847 846 1 1 30 0 500 711739 469
07dec2015 -0.48% 2 .007341 1587 03dec2015 1634 848 846 2 1 30 0 500 711740 469
08dec2015 -0.94% 2 -.0108091 1587 03dec2015 1634 849 846 3 1 30 0 500 711741 469
How could I make the dynamic prediction I need?
Thank you in advance
Rosa
Related Posts with Dynamic prediction after GARCH
Generating unique ID variable from numeric and categorical variablesHello all, I am new to forum and I have a question about generating new IDs. Thanks for your patienc…
reshape commandhi, I need to reshape a huge data set wide to long (2557 binary variables and 1.6 million observatio…
Finite mixture model vs. Latent class modelHi scholars, Would you please explain the difference between Finite mixture model and Latent class …
Using Multiple Imputations in Propensity Score MatchingHello everyone, I want to see the impact of a treatment, hence, using the propensity score matching…
Hosmer–Lemeshow test for large data setsHosmer–Lemeshow test for large data sets gives low p value despite perfectly fitting model. Increasi…
Subscribe to:
Post Comments (Atom)
0 Response to Dynamic prediction after GARCH
Post a Comment