I have financial price time series data and it has some noisy (probably erroneous) data and it also has gaps. As a first step I want apply a Tukey median smoother on points that do exist so I can study the residuals to decide if the point is an outlier. So far approach looks very promising on full sets of data that are noisy.

When applying smoothing to data with gaps how should I treat the points that are adjacent to the gaps. Simple approaches could be to (a) concatenate the data and treat as if no gaps (b) treat each section of data between gaps on its own and apply standard Tukey end point methods. Is there an approach that improves on these simple methods?

  • $\begingroup$ A comment on this having tried different approaches using linear interpolation on raw data before smoothing has worked best for what I am doing. Which is using the smooth to create residuals that can be analysed for outliers. By doing this if I remove outliers and rerun iteratively the process reaches a more stable conclusion. $\endgroup$ – James65 Feb 22 '17 at 6:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.