Discussion: View Thread

This message has been cross posted to the following eGroups: Young Professionals Group and Statistical Consulting Section .
-------------------------------------------
Hello,

I am working on a project that requires me to quantify seasonality and volatility in time series data. My data on state wide basis has some seasonality in it and is very volatile. In the long run, I would want to perform clustering on my variables so that states with common seasonal components would be grouped together as well as those that are volatile. Is there a paper/article that could be of help to me that you know of or if you have any idea as to how I should go about it, I would be glad to hear your opinion.

Thanks.
-------------------------------------------
Sandra Addo
seaddo@uga.edu

I don't know if you mean that a particular time series is both volatile and seasonal or that some are seasonal and others volatile.  Also what exactly do you mean by volatile?  Is it "high" variability over short periods of time?  I am not sure exactly how volatility is measured.  But with respect to seasonality, spectral analysis could be used (if time series is stantionary) to determine periodic components.  You could then fir a seasonal ARIMA model or a harmonic regression model and statistically test for the significance of the seasonal componet.  If you have several concurrent time series that appear seasonal spectral analysis will tell you whether or not they have common seasonal components.
-------------------------------------------
Michael Chernick
Director of Biostatistical Services
Lankenau Institute for Medical Research
-------------------------------------------