Notes on Fourier analysis
These notes on Fourier analysis are just a sketch, but are a good starting point for developing an analysis using autocorrelation and Fourier components to search for structure in the GCP data.
Date: Wed, 2 May 2001 11:04:40 -0400 (EDT) From: rdnelson [firstname.lastname@example.org] To: Peter Bancel [email@example.com] Subject: Re: gcp suggestion and Medreg update On Mon, 30 Apr 2001, Peter Bancel wrote: > I've enjoyed getting the occaional news about the GCP analysis. One line of > analysis that I have been thinking about is to look for signs of local > effects via a simple Fourier analysis. If local consciousness affects the > data, then there may be a signal that varies with the gross charater of > local conscious activity - such as when (most) people are awake or asleep. > It would be interesting to see if a data subset generated on one continent, > say, shows a bump on a 24 hour cycle. Similarly, one could look for a 7-day > periodicity since there is definitely a collective mood that changes as we > cycle through a week. Something like 15-minute data bins would probably > suffice and one could look over a few months in order to have enough data > points. If you or DM have a database set up to do this you could check it > with very little effort with a FFT. I would love to see such an analysis, and actually have something like it on the "to do" page of hopeful notions on the GCP site. Such things all require someone to be the doer, of course, so they wait 'til someone comes along. What I can do is help with discussion of sensible ways to treat the data matrix, how and whether to compress the raw data to minutes or other blocks, what to do about missing data, information about how to access the data, the processing scripts, etc. I will add your more specific suggestion to the to-do list, and hope someone is inspired to take it up.