Time Series Analysis

TommyGunne

Long time lurker in this forum. Don't post much here. If I have any input people tend to beat me to the ball!

Couldn't think of a more suitable forum, but I'm sure there is one as this doesn't seem to quite fit in. Not exactly a mathematical question, just looking for advice on something vaguely related.

I'm doing a time series project at the moment. I have to analyse two data sets, one of which has to have a seasonal/cyclical component. Analysing time series is EZ. The problem lies in finding easily available data for a seasonal time series. I don't particularly want to use an economic time series with cyclical components as I don't hugely believe in cyclicality in economics (function of other things unrelated to time) and now I'm just trying to find a data set that suits my purposes.

Any suggestions? Any input is much appreciated.

Also, if there is a more suitable forum, just let me know.

Thanks in advance for any help.

Fremen

I have a rather embarrasingly vague idea of what a time series is, but I gather that you want some data which changes cyclically with time?

My first thought was oil or gas prices, which tend to rise in Winter, but if you don't want to deal with economic data, why not try weather patterns? Average temperature or rainfall would show huge seasonal variation and would likely be very easy to obtain data for.

Edit:
link
http://www.met.ie/climate/monthly-data.asp?Num=76

If you need more detail, it looks as if they collect daily data, so maybe if you wrote them an e-mail...

TommyGunne

Stuff like this would do very nicely. Thank you for the input. Any more suggestions are welcome also! I'll follow this up now.

Also, not a definition, but a general understanding of what a time series is is basically a function that is plotted against time. However, this would include a lot of series that are straight line/direct functions etc. A time series is usually take to mean a series with some element of uncertainty, and hence one that is usually statistically analysed rather than directly mathematically (don't give me crap about stats being maths cos you know what I mean!) Usually taken to mean a series with some element of white noise/random uncertainty, also often includes a trend and seasonality and other stuff. The examples you came up with would be perfect examples of what are usually analysed using time series methods.

Also, I'd really love to get something really original, as there are 60 people in my class, and from what I gather, most people will be using meteorological series.

Fremen

When you say "cyclical data", my mind instantly jumped to weather. In all likelihood, that's what will happen with your classmates too. You could try something over a much, much shorter timescale.

Signal processing and electronic engineering in general is full of stuff like that. The engineering forum will be able to give you better examples than I can.

TommyGunne

Regarding oil or gas prices, I assume you mean prices seen in homes? Cos commercial and crude prices are not seasonal as this means that there would be arbitrage opportunities? Getting a significant data set of oil/gas prices seen in homes would be quite tricky I would have thought....

Fremen

There are indications that it's cyclical, but it's not like you can set your watch to the cycle. Seems doubtful that the trend would be strong enough to arbitrage from. Link to a paper:
http://cat.inist.fr/?aModele=afficheN&cpsidt=15885269

the disclaimer being that I haven't read it and from the abstract it may in fact be propoganda from OPEC.

Anyway, my original point was that a harsh winter will typically drive up the price of heating oil.

Fremen

Walking home listening to the beatles it occured to me that it might be interesting to get some oscilloscope readings from whatever mp3s you have lying around...

MathsManiac

Do you know anyone who owns a shop? Daily turnover might do the job. There's presumably be a weekly cycle involved and perhaps some longer phases too.

TommyGunne

Yeah I ended up doing sales cycles. Cheers for that. Obv weekly and annual seasonality in the model.