Persistency Filter (%)

Persistency Filter (%)

The default value is 2 (e.g., 2%) and the valid range is from 0 to 100.

As part of the spectrum analysis process, a test is performed on each cycle identified in the data to determine the persistency and the consistency of the cycle. Persistency is a measure of the degree to which the cycle is present in all of the sample data and Consistency is a measure of how well the cycle phase and amplitude match as the cycle progresses through the data. A combined score is created and this score can be used to filter out the least relevant cycles.

Assigning a value of 2 to the Persistency Filter parameter will filter out those cycles whose score falls below the 2% threshold in terms of Persistency and Consistency. We have found that a Persistency Filter range of between 2 and 10 works best for most financial time series. Financial time series are inherently noisy and the phase and amplitude of cycles is constantly shifting as you move forward in the data stream. As such, using a Persistency Filter value that is too high may indeed strip out meaningful cycles from the data.

For more on this topic, see Cycle Analysis