4
$\begingroup$

Let's say I have a database to track my customers.

On any given day, a customer is active or non-active, and I can discover this in real-time by querying the database. So, right now, I can easily find out that I have X active customers.

What I can't do is discover how many customers I had on July 5, 2014, for example. There is no way to reconstruct a particular point in time, to find out this data point for any given moment in the past. (Not uncommon -- it would be fairly rare, actually, for a database package to do this. Databases generally know about their state at the current moment only.)

To solve this, I create a job that queries the database every night at midnight, captures this data point, then writes to a file with a date. So, I end up with a series of timestamped numbers which represent this particular point of data at a regular interval over a period of time:

2014-07-05: 82
2014-07-06: 84
2014-07-07: 81
2014-07-08: 76
[etc.]

These numbers represent this data point, trending over time. This is a historical, indisputable record of a single data point at a single, specified moment in time.

Is there a name for this particular aspect/paradigm on this number? What would I call this data set, collectively?

$\endgroup$
4
$\begingroup$

You could call it a time series or a time-series variable (though most people neglect the hyphen). You could also use the more general term of longitudinal data; time-series data entail one type of longitudinal data. (If the measurements were taken at only two points, this would hardly constitute a time series and would have to be analyzed with other methods besides time-series methods.)

Analysts typically seek to answer two types of questions using time-series data. One is "to what extent can we explain the changes that occur?" -- as by creating a model of that single variable. ARIMA modeling is an example of such an approach. A second is, "To what extent can we explain changes in one such variable as a function of one or more others?"

(A bit more on terminology: while the series of numbers is a variable, the term "data point" applies only to any individual number, such as the 82. And then "data set" would usually refer to a group of such variables that might be analyzed together.)

$\endgroup$
  • $\begingroup$ So, a single captured value is a "data point"? A "time series" might be described as "a collection of N data points"? The answer below used the term "observation." How does that relate? $\endgroup$ – Deane Dec 27 '16 at 14:27
  • $\begingroup$ Yes to the 1st 2 questions. I'll hold off on paraphrasing IrishStat's answer, but usually observation simply means data point. $\endgroup$ – rolando2 Dec 27 '16 at 20:56
3
$\begingroup$

It is called a time series observation. This is a bit of an oxymoron or a contradiction in terms because it is never truly observed although it is called an observation. It is a collection of transactions .. your customer data (which can be observed) but it itself is never truly observed. We observe transactions and then bucket transactions into time or space intervals and then analyse the collective to make predictions and even more importantly detect innovation.

$\endgroup$
  • $\begingroup$ For the record, your answer was perfectly good, but I can only pick one as correct and both answers were basically the same... $\endgroup$ – Deane Dec 27 '16 at 13:58
  • $\begingroup$ If you want to learn more about time series search on user 3382 as it is the only subject I know . regards $\endgroup$ – IrishStat Dec 27 '16 at 15:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.