I think about having data such as:

Time   1   2   3   4   5   6   7   8   9 ... n

Set1  12  13  13  23  26  27  40  13  13
Set2   2   3   4   2   4   3  10  12  11 
Setn ...

All datasets are equivalent in terms of the meaning what a sudden (relative) change in values indicate. Means when in set1 is a change from 13 to 23 it indicates the same as when there is a change from 3 to 10 in set2. It only depends on the relative change of the actual t0 to the average of the past x periods.

My question is what is the best way to quantify such change? My goal is to have a value in each period for each dataset that indicates the change. This value should be comparable to the other datasets. Means when I have a change from 13 to 23 the value should approximately similar to a change from 3 to 10.

Would appreciate any advice!

Thanks and Best :)


1 Answer 1


To capture the relative effect i would go for x_t - x_t-1 / x_t-1 If you want the same result in your example x_t-1 = 13 x_t = 23 & z_t-1 = 3 z_t = 13 you could just go for the difference x_t - x_t-1

And if you want to put everything to respect of some stylized statistic, you could just change the denominator x_t-1 to t0 (t0 being median, mean, whatever you like)

  • $\begingroup$ Thank you. I also thought about something like this, maybe using not just the t-1s period but the average of the last x periods. This seems to be a very straight forward approach, does it have a name? :D I'm sure it has but I don't know it right now :D $\endgroup$
    – Marl
    Aug 9, 2019 at 11:57
  • $\begingroup$ Try to make the formula readable either by using math writing or using parentheses. x_t - x_t-1 / x_t-1 = x_t - x_t - (1/x_t) - 1 = -(1+1/x_t) $\endgroup$
    – user158565
    Aug 9, 2019 at 15:44

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.