I was wondering what time series model would forecast future values in such a static but decreasing manner? ( I'm talking about the values around where the circle appears, as apparently, the data set that I have does that, judging from the graph, but I'm a beginner and not sure what model would do that? though I think it should be a simple one). I just need reasonable ideas :)



  • $\begingroup$ the y axis shows the size of some population at a given point in time, if that is of any help $\endgroup$
    – Sarunas
    Sep 23, 2013 at 11:10
  • $\begingroup$ Could you please explain it a bit more? I still do not understand what's the problem... $\endgroup$
    – a.desantos
    Sep 23, 2013 at 11:28
  • $\begingroup$ Post your actual historical data and the future values and I will try and reverse engineer the equation. $\endgroup$
    – IrishStat
    Sep 25, 2013 at 21:12

1 Answer 1


If you have available the data set of the predicted values (and not just the graph), then you could apply "reverse-engineering" by trial-and-error, i.e. try to "estimate" a function from this data, treating them as random although they are not, by applying standard time-series approaches. In principle you would eventually find the exact forecasting equation.

But generally speaking, the fact that for a specific interval (the circled values) the forecasting function appears to behave as a "~linear function with structural shifts" can not, on its own, help "guess" the function, because, in other parts of the graph the function exhibits different behavior.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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