CODEWITHSUNDEEP
javascriptmathstatistics
sdailey
sdailey

Reputation: 2030

Time-series Statistics (such as correlation, Fourier transform)

[Note: I am a PHP/Javascript/Web Developer, and my weakpoint is math/stats]

I have written code that fetches time-series objects. (No problem here.) And I have a graphing library all linked up, ready to plot any given series.

How can I apply pattern-finding operations (such as a cross correlation function or perhaps a Fourier Transform) to the time series of interest? I've looked into jStat (http://www.jstat.org/) which is a port of R, but the API documentation is really non-existent. Furthermore, I'm pretty inexperienced in Mathematica and R.

Where should I start?

I haven't a clue how to reduce functions like the one found in this article into a useful operation. Cross Correlation between a pair of [time series] (it is math oriented): http://en.wikipedia.org/wiki/Cross-correlation

Upvotes: 2

Views: 1631

Answers (1)

sdailey
sdailey

Reputation: 2030

This old article from 1996 proved to be the best starting point. It includes a straightforward definition of cross correlation and auto correlation, as well as some portable c-code. The resulting JS code had no dependencies and left me with an actual understanding of what was being done.

Although the original question is no longer important to me, please share any additional resources on how to get others started if you know!

http://paulbourke.net/miscellaneous/correlate/

Upvotes: 3

Related Questions