Talk:Time series

Question about terminology
This topic is inconsistent about hyphenating the term "time series" as a standalone noun (time series/time-series) and as a modifier (time series analysis/time-series analysis). Can someone decide?

Proposals for restructuring the article
I am wondering how to make the article more accessible to the non-expert.

I am also thinking that we need a succinct introduction to time-series with some good motivating examples before we plunge into the detail. I will think a little about how to do this. --Михал Орела (talk) 16:30, 19 October 2008 (UTC)

Image of time series
"Image is everything"!

Well, of course, it is not. But in our modern world it helps if we can attract and inform at the same time. I have chosen a nice image from Wikimedia Commons to illustrate the article. I think it looks good. I think it gives some idea of what a time series might look like and what one might want to do with it. The author has contributed a detailled commentary: “The data is 1000 points, with a trend of 1-in-100, with random normal noise of SD 10 superimposed. The red-line is the same data but averaged every 10 points. The blue line is every 100 points.

The r2 fit for the raw data is 0.08; for the 10-pt-filtered, 0.57; for 100-pt-filtered, 0.97.

For all series, the least squares fit line is virtually the same, with a slope of 0.01, as expected.

Ignoring autocorrelation, a confidence limit for the fit line is [0.0082, 0.0127] for the raw data (which include 0.01, as it should). For the 10-pt-filtered the limits are slightly narrower at [0.0084, 0.0125] and for the 100pt-filtering the limits are again slightly narrower.

So what does that all mean?

for the raw data, the simple trend line explains almost none of the variance of the time series (only 8%). for the 100-pt filtering, the trend line explains almost all of the data (97%). nonetheless, the trend lines are almost identical as are the confidence levels.

The time series are, of course, very closely related: the same except for the filtering. This shows that a low r2 value should not be interpreted as evidence of lack of trend.”

and also provided source code. --Михал Орела (talk) 07:26, 20 October 2008 (UTC)

62.56.97.111 (talk) 23:29, 18 November 2007 (UTC) Does need examples towards the beginning of the article

what is time series data and give examples

// Changed the one section to Applied Time Series, which is IMO better than "Industry" // Added Multiscale to the introduction to reflect recent developments // THIS IS NOT part of economics, Time Series is STATISTICS. Econometrics is the topic that should be "part" of economics, the two are linked but SEPERATE. —Preceding unsigned comment added by 82.32.9.240 (talk) 11:17, 16 September 2007 (UTC)

Free software
Nearly all the links to software are for pay-for stuff - could not some links to free software be included also? I am not an expert on relevant software myself.

What does the alpha term mean in the time series, and what is the significance of Y? Is it the sum of all of the terms? Why would you want to do that? —Preceding unsigned comment added by 192.85.47.12 (talk) 19:13, 8 January 2008 (UTC)

Will do —Preceding unsigned comment added by 62.30.156.106 (talk) 18:31, 23 January 2008 (UTC)

I nixed Mathematica, if this is a thing I'll nix a couple more. It's a bit strange to me because I really don't hear about the paying ones at all. Reason being if it's proprietary, then you don't get the community writing code and packages for it, so it becomes much less useful. Ie for what I do in time series people don't talk about Mathematica at all full stop. It's all free open source stuff like Python and R. Python has TF, Keras, and PyTorch, Mathematica has nothing. So I'm right there with you guys on that. It's not that I'm against proprietary software, but for the stuff that I do, the only proprietary software I hear about is CUDA, and that's something related but different. Alcibiades979 (talk) 00:58, 30 July 2019 (UTC)

Suggestion - Contrast Time Series with it's alternative(if there is one) for clarity.
Is there an alternative representation of data other than Time series? Frequency Series maybe? If someone could mention an alternative and contrast it in the introduction it would greatly clarify what Time-series are, besides being a fancy name for a bunch of data points with time as one of the dimensions.Krymson (talk) 23:38, 2 February 2008 (UTC)


 * Have added something on problems that are not "time series". Melcombe (talk) 14:46, 15 February 2008 (UTC)

External Link DonQuixote
I have added a link to the home page of DonQuixote time series software. It is one of the few free time series software available. It is probably the only free software which includes free GPL source code in C++. It is a quite extensive package with more tha 100000 lines of C++ source code. The linked web site does not include anything else than the homepage of the free time series software. —Preceding unsigned comment added by Truswalu (talk • contribs) 14:32, 29 March 2008 (UTC)
 * Please read the external link guidelines I've provided you; specifically, links requiring registration to view the 'content'. Thanks.  Kuru  talk  15:10, 29 March 2008 (UTC)

References to the (usual) literature
It is usually a good idea to cite the published literature (in addition to web pages). The Time series field is well-established. I will add in the appropriate section now and list one text to get started.--Михал Орела (talk) 14:11, 19 October 2008 (UTC)

Properties section or just description section
I want to list all possible properties of a time series: volatile, persistent, stationary vs. explosive, mean reverting vs. structure break, and so on. Also a list of possible properties of several time series: related or not, cointegrated or not, and so on. Jackzhp (talk) 20:36, 18 May 2011 (UTC)

Evenly and unevenly spaced time series, wrong examples
The article again says "a time series is a sequence of data points, measured typically at successive times spaced at uniform time intervals. Examples of time series are the daily closing value of the Dow Jones index ". Well the Dow Jones Index is exactly an example of a time series that is NOT evenly spaced: It is published only on trading days, so the interval between values is often 24 hours but also very often 23 or 25 hours (daylight saving changes) or several days (weekends, holidays). Generally, time series analysis processing has often to do with real world events like transactions at a stock exchange. A live feed of transactions about a certain stock will deliver values in random timely order: There are seconds were several transactions occur and there are hours were nothing happens. I would also remove the restriction that a time series is a measurement, because transactions at an electronic exchange for instance are computed by the computer there and communicated afterwards. I suggest a time series definition that says that defines it as an sequence of values each having a time and ordered by time. — Preceding unsigned comment added by Thulka (talk • contribs) 12:22, 12 January 2012 (UTC) The second example, yearly flow is not evenly spaced either, as gregorian calendar years to not have constant length due to leap year and second effects. The example I brought in here with the audio sampling was a correct example.

The article on unevenly space time series is good, but the insight that nearly all natural observations are unevenly spaced is missing from this article.

Time series or time sequence?
What is the exact name of this notion time series or time sequence(s)? Is not a sequence? This aspect is useful to be clarified.--5.2.200.163 (talk) 14:56, 25 September 2017 (UTC)


 * You could argue that the concept should logically be called a "time sequence", because series in mathematics implies that the terms are summed. But "time series" (however illogical it may be) is the common name for this concept, so our article titles policy says that is what the article should be called. Gandalf61 (talk) 15:05, 25 September 2017 (UTC)
 * I think that some historical aspects about the how the name time series in current form has becomed what it is and who has named it this way instead of time sequence should be added to the article. (I also notice that this rather improper name has been passively(?) adopted by translation in languages other English).--5.2.200.163 (talk) 09:23, 3 October 2017 (UTC)

Example software
The article currently contains this list:


 * 1) CRAN supplementary statistics package for R
 * 2) Analysis and Forecasting with Weka
 * 3) Predictive modeling with GMDH Shell
 * 4) Functions and Modeling in the Wolfram Language
 * 5) Time Series Objects in MATLAB
 * 6) SAS/ETS in SAS software * Expert Modeler in IBM SPSS Statistics and IBM SPSS Modeler
 * 7) Automatic Time series Forecasting with LDT * EViews is a statistical package for Windows, used mainly for time-series oriented econometric analysis.
 * 8) bayesloop: Probabilistic programming framework that facilitates objective model selection for time-varying parameter models
 * 9) Slycat Web-based ensemble analysis and visualization platform, created at Sandia National Laboratoriesl
 * 10) Seglearn: open source python package and scikit-learn extension for machine learning with time series and sequence data

All the references here are primary (i.e. not independent from the package), and some are not notable (in Wikipedia terms). Though I am familiar with some of the packages here, I doubt for others that they are particularly known outside of the selected field (and the sources certainly don't show it). Those that do not have their own wikipedia article should have sources which are both independent (secondary) and reliable, showing that they should be included in this list (basically, 'notable' use of the functions named in this article). Such sources could also include reliable comparisons/reviews of the packages (where that could be a general source for the section).

I've asked today for independent sourcing to be able to show that packages warrant being included in this list. In about a month any that do not have any independent sourcing showing that they belong in this list, AND do not have their own article to show that they are notable should be cleared from this list. --Dirk Beetstra T C 11:05, 28 September 2018 (UTC)