Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 2 down vote favorite

I am studying a complex system. My goal is to understand the impact of a spreading accident (which saturates the filter (sand and plant) and creates puddles). Puddles decrease the efficiency of the system (depollution of effluent, and moisture content). The impact on the efficiency will decrease with time (more or less depending on the season and meteorological situation).

I want to develop a model of how long the incident will impact on the pilote. For that I think I should use time series analysis.

My question is:

  • Can I use time series analysis on short time scales (e.g., with 5 to 30 measurements?
  • or can I compare the effluent volume with the parameter that I want to study, with a decalage which should not be regular because of its depending with the effluent volume injected?
share|improve this question

1 Answer

up vote 3 down vote

Time series nlysis works best if the series is long enough to detect periodicity or trends. It usually will not work very well on short series. However if you have several short series and you are looking for differences among them then you should be doing longitudinal data analysis such as repeated measures analysis of variance.

share|improve this answer
Thanks, one more question, to use longitudinal data analysis, shouldn't I independant variable? because I can cut my data in différent periode, especially each time that there was a incident and compared them each-other but I can justify the homoscedasticity. Moreover, my data do not have normal distribution, and until now I did not find a way to intergrate them in order to have a normale distribution, so I shoul use non-parametric test which are less efficient!? – Isabelle Mazerie Jun 7 '12 at 12:52
My answer to non-normality is that it depends on how large a departure from normality your data is. For mild departures the standard methods will still work pretty well. For extreme departure you can use nonparamtric methods or try the bootstrap. – Michael Chernick Jun 7 '12 at 13:35
What do you mean by "mild departures or "extreme departures"? (I read a lot french texxt on stat, so I do not have all the meaning of english vocabulary, and I admit that I am not sure of the translation! and thanks again, this will help me a lot. – Isabelle Mazerie Jun 7 '12 at 15:05
@IsabelleMazerie Mild means slight. So i am saying that if the distribution is only slightly skewed for example the standard inference assuming normality will work okay. – Michael Chernick Jun 7 '12 at 15:07
ok, I think I get it! Thank you! – Isabelle Mazerie Jun 7 '12 at 21:28

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.