I have some time series to analyze.
Given the domain the data is coming from -
- Time series is supposed to have some fluctuations.
- A regular periodicity might not be present at all in some cases. There might be some irregular periods of droughts (no fluctuations happening at all)
- These fluctuations may be a part of an overall down/up trend.
I am trying to avoid modeling techniques like ARIMA etc. since I am only interested in knowing the following features for each one of them:
- Average amplitude of fluctuations.
- Average time period of fluctuations (how long it takes for values to rise and fall back to almost same level?).
- Average frequency of fluctuations. After what period do these fluctuations occur?
Following is what some of the data looks like:
The approach I am taking is to -
First build some sort of annotation on the time-axis (e.g. flat, increasing, decreasing)
Then based on these tags study further the patterns to answer the above questions.
In case there is an overall up/down trend in the series I am de-trending it by removing mean/linear-fit, etc.
I was wondering if there is any other approach or technique to answer the above mentioned questions for my data.