Hi there;
I am doing my PhD research to invistigate the effect of wavelet denoising on stock market volatility using all standard garch models(GARCH,EG,AP,TG,CG). I already de-noised the return sereis using the wavelet transform and that already done on Matlab trying to estimate the models in eviews. What iam having is persistence level (alpha + beta) greater than one in GARCH model for the denoised return sereis and lower than one for the contaminated (original) return series. I beleive that this represents violating of the standard GARCH condition and therefore the existence of explosive nature of conditional volatility. What i cannot understand is why for all models the out-of-sample forecasting error decreased for the denoised data. In other words, why the in sample estimation is very bad and the out-of sample is good with such improvements in forecasting? Please find attached and note that again i denoised thre return series and for all estimations i get (alpha) estimate is lower for the standard garch models and higher for denoised-series based GARCH models.
looking forward to hearing your reply as soon as possible.
Please noe that
aexorg denotes that the workfile with garch estimaitons for AEX index of original return series.
aexhard denotes that the workfile with garch estimaitons for AEX index of wavelet denoised return sereis using hard threshold.
aexhard denotes that the workfile with garch estimaitons for AEX index of wavelet denoised return sereis using soft threshold.
GARCH Persistence & Return Series Wavelet De-noising
Moderators: EViews Gareth, EViews Moderator
-
- Posts: 13
- Joined: Thu Mar 27, 2014 9:17 am
GARCH Persistence & Return Series Wavelet De-noising
- Attachments
-
- omxsoft.WF1
- (87.54 KiB) Downloaded 261 times
-
- ARXHARD.WF1
- (87.55 KiB) Downloaded 234 times
-
- AEXORG.WF1
- (87.78 KiB) Downloaded 228 times
-
- Posts: 13
- Joined: Thu Mar 27, 2014 9:17 am
Return to “Econometric Discussions”
Who is online
Users browsing this forum: No registered users and 22 guests