1

I have 2 timeseries representing the energy expenditure of 2 group of patients. Data are sampled every hour. The resulting plots represent indeed the hourly pattern of the EE of the 2 groups.

Is there a statistical test in order to compare (and emphatize differences) between 2 hourly patterns?

Thanks!

gabboshow
  • 641
  • 6
  • 17
  • What kind of differences are of interest? – Silverfish Nov 03 '15 at 10:22
  • @Silverfish I created a more detailed question... please have a look at http://stats.stackexchange.com/questions/179947/statistical-differences-between-two-hourly-patterns I can't close this one... – gabboshow Nov 03 '15 at 10:25

2 Answers2

0

Coherence will indicate whether two series have the same frequencies. Phase shift will show what is the phase shift at these frequencies.

Aksakal
  • 55,939
  • 5
  • 90
  • 176
  • Would it work if I have only 24 data points (one each hour). – gabboshow Nov 02 '15 at 15:04
  • It depends on the pattern. It will certainly not work for any kind of pattern repeating daily or monthly. You can only detect frequencies between once in half hour and once a day. – Aksakal Nov 02 '15 at 15:16
  • Could you please have a look at this question? http://stats.stackexchange.com/questions/179947/statistical-differences-between-two-hourly-patterns I asked to close the current one since it is very unclear... – gabboshow Nov 03 '15 at 10:05
0

You could use Hodrick-Prescott filter to detect trends and cycles (daily or weekly) and compare among the two data series.

Davide C
  • 409
  • 1
  • 3
  • 9