Here's a question for you quants out there. Given a specific financial markets data set, in real time how do you choose your forecasting technique/method/model? Intuition, experience, mathematical rigour, or a combination of all three?

Hi pedro,The best way is the combination of all three. Intuition and experience along with mathematical rigour will always give best results.As far as mathematical technique/method is concerned you can proceed in the following sequence:1. find out the auto-correlation and partial auto-correlation to figure out the MA and AR lags.2. Depending upon the MA and AR components use the appropriate estimation technique. ( for example if you consider that the errors are normal iid then depending upon the MA and AR components and no. of data points use conditional or exact likelihood function)3. statistical test for the estimated parameters.of course you have to take care of seasonality component (where you experience and Intuition will help the most) etc.mpsingh

OK then, I have another question for you; do quants who work for traders do much forecasting as such (i.e. ARIMA, EGARCH) or is it mainly asset pricing of financial instruments?

Why forecasting is so important?

You tell me

OK, I've found a useful thread in the FAQs Forum GARCH

pedro, just out of interest, how old are you?

I never give out that sort of information on a first date - and sometimes only to close friends and family

QuoteOriginally posted by: pedro666Here's a question for you quants out there. Given a specific financial markets data set, in real time how do you choose your forecasting technique/method/model? Intuition, experience, mathematical rigour, or a combination of all three?mpsingh answer is a good starting point. You want to try first linear model for your timeseries. In general forecasting is more an art rather than a science. In a nutshell the following steps are common:1. Decompose your time series into: trend, cyclical (seasonal or low frequency component) and stationary timeseries. 2. Find a model for each component. Here your intuition about market structure or other economical wisdom might be usefull3. Split your data set into 2 parts, use part 1 to estimate the parameters and part 2 to do out of sample test.

pedro666,What's your gender?Bashir

QuoteOriginally posted by: bashirfpedro666,What's your gender?BashirMale. Why do you ask?

QuoteOriginally posted by: ErrrbQuote2. Find a model for each component. Here your intuition about market structure or other economical wisdom might be usefullCan you give me a real life example?