0

I am working on GPS data and I have plotted the speed distribution across a time frame for 3 different moving objects. Data length is not the same for my 3 entities. i.e object 1 has 4500 rows and object 2 has 2700 for each attribute. moreover, time is in seconds, starting from the first measurement.

my code is as follows:

plot(object1$speed)
plot(object2$speed)

which gives me the following graphs:

[enter image description here1

[enter image description here2

I am trying to find a way to compare them. that is measuring similarity. I am also trying to see if there is a similarity in some particular part of the distribution if not all. I do not want to correlate the distributions as this will give me just one value. I am trying to find similarities in all or part of the sequence. does anyone know how to do this?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
  • correlation with a sliding window? – OganM Aug 17 '16 at 20:55
  • 1
    Why not just graph `object1$speed - object2$speed` where times are equal? – Chris Aug 17 '16 at 19:38
  • 1
    Perhaps you should read [this paper](https://www.cs.uic.edu/~tdang/file/TimeExplorer.pdf) by Dang and Wilkinson on comparing time series by a set of 11 common features. The latter author is the progenitor of the "grammar of graphics" as espoused in `ggplot2` or `plotly`. –  Aug 17 '16 at 19:53
  • @Chris. do you mean just putting them on the same plot? times may not be equal at all. what I am really trying to do is find similarity in their speed distribution pattern. –  Aug 17 '16 at 19:39
  • [Here](http://stats.stackexchange.com/questions/19103/how-to-statistically-compare-two-time-series) is a similar post to get you started – bdeonovic Aug 20 '16 at 17:02
  • 2
    Please explain what you mean by the plots of two objects to be "similar." (There are many possible different meanings, depending on what aspects of their trajectories you are studying.) – whuber Aug 21 '16 at 15:16

0 Answers0