0
$\begingroup$

I'm have a time series that is dependent on a large number of other timeseries, but these dependent timeseries don't add up to the main one, as I don't have the full population of these dependent timeseries, only a sample.

All of these dependent time series are likely to be different and they are unlikely to be randomly sampled from the population.

I was thinking of forming a linear combination of these dependent timeseries to try and then use chi-sq minimisation to find the values of the co-efficients.

As the superposition could have about 60 coefficients (timeseries), I think the problem could be quite degenerate.

My question is: does this sound reasonable, and what problems am I likely to run into, or is there a more powerful way of doing this?

Improve this question
$\endgroup$
2
  • $\begingroup$ Could you perhaps explain more: what data you have; whether you have the independent time series or not; whether each dependent time series is a superposition of the same set of independent time series or not? $\endgroup$
    – conjectures
    Commented Feb 26, 2015 at 12:44
  • $\begingroup$ It's a series of power readings over time, we know the input into the building and the values at certain sections in the building, but I'm trying to gauge the importance of each timeseries in relation to it's effect on the total. $\endgroup$
    – James
    Commented Feb 26, 2015 at 16:09

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.