Conversation

Notices

1. Muninn 🍃 (muninn@social.coop)'s status on Thursday, 21-Dec-2017 13:25:14 EST Muninn 🍃

   Can ppl boost this until somebody who knows #stats or #dataviz etc sees it?

   I have a time series of numbers, sampled irregularly, typically between 1 and 6 samples a day. On a line graph, there's considerable jitter. If I want to smooth jitter and see trends on e.g. a 36-hour granularity level, but also not lose outliers in the smoothing, what's the best approach?

   My idea: average samples in arbitrary 12-hour periods (am/pm), then apply a 36-hour moving average. But I don't know this stuff.

   1. arunisaac (arunisaac@social.systemreboot.net)'s status on Thursday, 21-Dec-2017 16:10:58 EST arunisaac

      in reply to
      Maybe, plot the average along with minimum and maximum. Or, you could plot the average along with the standard deviation. Either of these techniques should capture any anomalous deviations from the average value.
      
      You could use an error bar plot to visualize this.
      http://gnuplot.sourceforge.net/demo/errorbars.html