Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

First of all, I'm sorry I'm a complete statistics iliterate, I promise I'm going to get into it after the summer (I need them for my job actually). Because of this I'm not sure about the name of what I'm asking about, I will explain and you tell me what I have to research, study and learn. Although it could look like I'm asking something about programming... I'm not, please read till the end.

I have an application that stores in a table the amount of time that takes a remote server to reply a "hello" every minute. So, basically I have the date and time of the request and the delay.I can open this data and show a nice chart using Flot.

The problem is, when I try to open in the chart all the meassures in the last month, takes a lot of time because there are thousands of elements. I think, what I should do is get a kind of "average" where from those thousand of points, I could get 100 that represent what has happened during the month. I know I will lose precision, but it doesn't matter because the user will be able to do zoom to a specific range.

So, what mathematical function/technique do I have to use to make that reduction?

Thanks a million.

share|cite|improve this question
It depends a bit on how you'd measure how well this subset of points "represents" the entire set of points, but for many practical purposes it might be sufficient to simply select $100$ random points (with the same selection probability for all points)? – joriki Jun 27 '11 at 13:25

You have several choices. You can just make a histogram, which shows how many times the response took 0-1 sec, 1-2 sec, and so on (choose your bin size to fit the data). You can report a mean response time, but this risks losing the important information that a few requests took much longer. You can report the 10 percentiles, which is essentially a histogram. You can calculate mean and sigma, but that tends to make people think the data is Gaussian.

share|cite|improve this answer
I thought the question was how to select $100$ "representative" points from the given points. – joriki Jun 27 '11 at 13:50
@joriki: I thought that was an approach, which has merit. I suggested some others. I hope one or the other is useful. – Ross Millikan Jun 27 '11 at 13:53

Compute the centils, it will be enough to know them to understand how your datas are, and it will only be 100 values to "load".

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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