Comparing R/S, Modified R/S, Wavelet, and DFA Tests for Long Memory
Summary
The document concerns detecting long-term memory in interest-rate data, including yield spreads. The question raises rescaled range analysis, Lo’s modified R/S test and its critical values, and whether wavelet methods might be more appropriate. The answer points to a later quantitative portfolio trading book as a source that reviews theoretical and empirical research on the fractal market hypothesis.
That source is described as introducing and comparing R/S analysis, modified R/S, wavelet analysis, and detrended fluctuation analysis, with simulations used to examine their efficiency. The exchange thus offers a route to broader method comparisons rather than resolving which test is best for a particular series. It does not report empirical findings about interest rates, assess the current status of Lo’s critical values, or give implementation guidance. Researchers would need to consult the referenced work and evaluate test assumptions and data-specific behavior before drawing conclusions about long memory.
Key ideas
- The question focuses on testing for long memory in interest-rate and yield-spread series.
- Methods mentioned include R/S, modified R/S, wavelets, and detrended fluctuation analysis.
- A later book is cited as reviewing empirical and theoretical research on the fractal market hypothesis.
- The cited work reportedly uses simulations to compare the efficiency of these methods.
- The exchange does not identify a universally preferred test or settle the status of Lo’s critical values.
Tags
Full text
# Longterm memory in interest rate data - R/S analysis # Longterm memory in interest rate data - R/S analysis Am currently investigating long term memory in interest rate data. The two sources I am using are Peters (1996), "Fractal Market Analysis" and an article published in the Journal of Fixed Income, "Investigating Long Memory in Yield Spreads" (S2009) by McCarthy, Pantalone & Li. I am unsure as to whether this type of question is appropriate for the forum but I would like to know if: 1) more recent research is available. An article reviewing the empirical studies conducted so far would be helpful. 2) if R/S (Rescaled Range) analysis is still applied today on financial time series (the book by Peters is > 20 years old) 3) or if wavelet-type analysis is considered more pertinent 4) and finally, in conducting R/S analysis, one can compute the V-stat to detect long term memory in the data. Lo (1991) published an article with an improved version of the test. The article contains a table of critical values. I would like to one if this is still most recent table to date? Thnx, ## Answer by Devin_Wang (score 1, accepted) https://quant.stackexchange.com/a/39249 There are two chapters of Baniel Bloch's A Practical Guide to Quantitative Portfolio Trading covering fractal market hypothesis. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2543802 The book is written in 2014, so it reviews a great number of more recent research in the last two decades, both theoretical and empirical. As for different methods used to detect long-term memory, R/S analysis, modified R/S analysis, wavelet, DFA, a basic introduction and comparison is available in the book, and Bloch also made a detailed simulation to test the efficiency of these methods. Hope this helps!
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.