Statistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
Statistics in finance, economics, engineering. Actuarial science. Econometrics. Operations research.enMon, 30 Mar 2015 07:03:15 GMTvBulletin60http://www.talkstats.com/images/misc/rss.pngStatistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
Correlated time series at different aggregation levels
http://www.talkstats.com/showthread.php/60194-Correlated-time-series-at-different-aggregation-levels?goto=newpost
Sat, 28 Mar 2015 17:10:05 GMTHi,
In my time series data I discovered the following behaviour. My time series are available in 500ms intervals and are about sensor data like...Hi,

In my time series data I discovered the following behaviour. My time series are available in 500ms intervals and are about sensor data like energy consumption, motor speed,...

When correlating those time series at the shortest time interval available, I don't get any correlation, because the signals of course have some outliers, noise, maybe lag. When I calculate the mean value for each minute they are correlated, because those aspects are negligible after the aggregation.

My question is now if there is some "theory" behind that "phenomenon"? Is there a scientific name or concept of that? Does it make sense to find out the aggregation level with the highest correlation and what does it tell me? It would be very helpful if someone could help me how I should classify this observation and if there it's a known and common challenge and how it is usually handled.

Thanks!
]]>Applied StatisticsJohnMeierhttp://www.talkstats.com/showthread.php/60194-Correlated-time-series-at-different-aggregation-levelsMultiple Correlation Coefficient for Time Series?
http://www.talkstats.com/showthread.php/60193-Multiple-Correlation-Coefficient-for-Time-Series?goto=newpost
Sat, 28 Mar 2015 17:08:59 GMTI am using Pearsons correlation coefficient to calculate the correlation between two time series. Now I would like to calculate the correlation...I am using Pearsons correlation coefficient to calculate the correlation between two time series. Now I would like to calculate the correlation between a set of time series A, B, C and the time series D. It should be a many-to-one correlation (A, B, C-->D).

Is there a common approach how to do this?

Is it valid to combine A, B and C by summing or convolution and then calculate the correlation between the aggregatef time series (ABC) and D?

Thanks for any help.
]]>Applied StatisticsJohnMeierhttp://www.talkstats.com/showthread.php/60193-Multiple-Correlation-Coefficient-for-Time-Series