I'm trying to do analysis on revenue data and due to circumstances beyond my pay grade, there are large spikes in the data around the time financial reports are due, followed by corresponding low values immediately after. Is there a preferred algorithm or methodology to compensate for this and get more meaningful data? It would seem that a simple moving average would be inappropriate for the task, given that I know there are discontinuities at time edges. Is there a better smoothing strategy?

  • 5
    $\begingroup$ Good answers to this question will need to know why would you want to smooth out these values. $\endgroup$ – Sycorax says Reinstate Monica Jul 9 '14 at 20:41
  • $\begingroup$ The company 'ships aggressively' to artificially inflate sales numbers for the current period. Items that would ordinarily ship in the first week of the next period are pulled forward to the current one. That does not accurately indicate customer behaviour. $\endgroup$ – Jeff Sacksteder Jul 10 '14 at 13:38
  • 1
    $\begingroup$ Well, quite obviously, it does reflect customer behavior... after your employer has "trained" the customer, probably involving end-of-quarter price incentives or something. The question really is what "meaningful data" in your question is. The data are the data. You could perhaps try some kind of smoothing to estimate possible sales without this effect and quantify the revenue lost through the end-of-quarter incentives - also the operational difficulties in the supply chain due to larger demand fluctuations. $\endgroup$ – Stephan Kolassa Jul 11 '14 at 9:27
  • $\begingroup$ Here are some methods for smoothing, you can study them for find your best choose or you should innovate a method, I guess you can use a smoothing just for times "around the time financial reports". $\endgroup$ – parvij Jul 15 '14 at 9:46

There are methods which use the knowledge of the point in time of the unusual event which leads to a window of response before and after the known event. These methods are called different things but one name is Dynamic Regression or Transfer Functions or armaX models.

  • $\begingroup$ if you want to post your data and the points of known interventions I would be glad to help $\endgroup$ – IrishStat Oct 1 '14 at 23:51
  • 1
    $\begingroup$ You sir, are hysterically over-qualified to answer my question. Thank you. $\endgroup$ – Jeff Sacksteder Oct 2 '14 at 15:33
  • $\begingroup$ @JeffSacksteder You are one funny guy .... I will try and reduce my knowledge/solution ideas and talk to you slowly. Please give me a call at my contact number .... regards $\endgroup$ – IrishStat Oct 2 '14 at 18:05
  • $\begingroup$ Or perhaps we can go to a chat room .. Please set one up and invite me .... $\endgroup$ – IrishStat Oct 2 '14 at 19:38

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.