0
$\begingroup$

I have a time series of data where the increment is every minute. In order to approximate AUC, I just compute the sum of the data values, since they would all be multiplied by 1 anyway per Riemann sum approximation of the integral.

I'm fairly confident that this means the data can be scrambled and still produce the correct AUC, but I was hoping to check here in case I'm missing anything silly.

Improve this question
$\endgroup$
5
  • $\begingroup$ What kind of data does your time series contain? $\endgroup$
    – m13op22
    May 19, 2020 at 19:46
  • $\begingroup$ I have the oxygen level for each minute of an experiment, and I need to report the integral of this. (I know this is not traditional AUC in terms of having an ROC curve, I just meant the area under the time series). $\endgroup$
    – pyhat32
    May 19, 2020 at 19:49
  • $\begingroup$ Okay, you want to change your terminology then. And then this question probably belongs on the computational science site. But an easy way to confirm that both methods produce the same results would be to sort the data by time and then integrate, and not sort and still integrate. $\endgroup$
    – m13op22
    May 19, 2020 at 19:54
  • 1
    $\begingroup$ what does the acronym AUC mean? $\endgroup$
    – Anton Menshov ♦
    May 19, 2020 at 20:38
  • $\begingroup$ The reasoning looks correct. I would cross check the overall normalization though. You may actually want the have the average and not the sum of the values. $\endgroup$
    – jolvi
    May 20, 2020 at 15:21

0

Reset to default

Your Answer

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