Thank u for jack's warmthly reply.

 

I am a new user in gretl and also a new baby in tseries.

Previous sample data was expanded variance from time to time . So It was hard to estimate garch model(i know now.. alpha + beta > 1 then convergence is not met). so I decided to include more data, but I don't know how many data is required to make properly garch model.

 

I was seen a book for few years ago about arima model. this model is required at least 50 data.

 

q1]Please tell me how many data is requiered to estimate garch model...

just includeing data until alpha + beta < 1 ??

 

q] In generaly estimating a model, 60% data is estimate model's a parameter and 40%'s data is confirm a model .. estimated model is properly estimated. I think. in time series this model confirming method is not met...

 

i will waiting for more warmthly answer again.