0
$\begingroup$

I'm new to data science. I'm trying to increase the time-series length for a special calculation. In the original time-series I have 20 weekly reports and I want to increase the amount of occurrences to 200. Is it ok just to use the range from the first date value to second date value between two neighbor dates?

For example if I have 1 for the first date and I have 5 for the second one, is it fine to populate the empty space between them with 2, 3, 4. Or do I need to use more advanced techniques here?

| improve this question | |
$\endgroup$
1
$\begingroup$

It depends on the further analysis that you want to perform. Increasing amount of occurrences can significantly effect the properties of your time series data, because you have only 10% of the data available.

In general, this is an interpolation question. There are fundamentally different statistical models that you can use to find the missing data. The Pandas dataframe.interpolate() is a good way to start.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thank you very much! $\endgroup$ – Дмитрий Сажнев Aug 22 '19 at 21:34
  • $\begingroup$ Actially I want to use Rachev Ratio formula for further calculations for this timeserie. And then I will drop interpolated data and will work with real timeserie. So it seems like interpolation will not have negative affect on the analysis. Thank you $\endgroup$ – Дмитрий Сажнев Aug 22 '19 at 21:41
  • $\begingroup$ Why not deriving your model based on the data itself without interpolation? $\endgroup$ – Mahdiar Aug 22 '19 at 21:55
  • $\begingroup$ because of the customer :) He doesn't want to accept that Rachev Ratio can not be calculated for such a small time-series. I know that sounds stupid but I need to finish that. @Mahdiar $\endgroup$ – Дмитрий Сажнев Aug 23 '19 at 8:01

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.