Paper: arXiv 2406.19406
Authors: Borko Stosic, Tatijana Stosic
Abstract
In this work we address the question of the Multifractal detrended cross-correlation analysis method that has been subject to some controversies since its inception almost two decades ago. To this end we propose several new options to deal with negative cross-covariance among two time series, that may serve to construct a more robust view of the multifractal spectrum among the series. We compare these novel options with the proposals already existing in the literature, and we provide fast code in C, R and Python for both new and the already existing proposals. We test different algorithms on synthetic series with an exact analytical solution, as well as on daily price series of ethanol and sugar in Brazil from 2010 to 2023.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 5.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper is densely mathematical, featuring extensive LaTeX formulas for multifractal spectrums, Legendre transforms, and derivations of scaling exponents, which warrants a high math score. Empirical rigor is moderate, as it provides C/R/Python code, tests on synthetic data with known analytical solutions, and real-world price data (ethanol/sugar), but lacks the deep, live backtesting environment typical of production-ready strategies.
Research Flowchart
flowchart TD A["Research Goal<br>Resolve controversies in MF-DXA<br>handling negative cross-covariances"] --> B["Methodology<br>Propose new options for MF-DXA"] B --> C["Data Sources"] C --> D["Computation<br>Implement fast code in C/R/Python"] C --> C1["Synthetic Series<br>(Exact analytical solution)"] C --> C2["Real Data<br>Ethanol & Sugar prices (2010-2023)"] D --> E["Analysis<br>Compare new vs existing methods"] E --> F["Key Outcomes<br>Robust Multifractal Spectrum<br>Validated algorithms<br>Published code"] style A fill:#e1f5fe style F fill:#e8f5e8
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes several new algorithmic options to handle negative cross‑covariance between two time series, aiming to provide a more robust assessment of the multifractal cross‑correlation spectrum.
Source sentence
To this end we propose several new options to deal with negative cross-covariance among two time series, that may serve to construct a more robust view of the multifractal spectrum among the series.
Assumes
No explicit assumptions found in the text.
Data
- Synthetic dataset used in [11] synthetic public
Source sentence
We test different algorithms on synthetic series with an exact analytical solution, as well as on daily price series of ethanol and sugar in Brazil from 2010 to 2023. We substantiate these proposals through a comprehensive assessment of a synthetic dataset used in [11], uncorrelated white noise series, and daily sugar and ethanol price data in Brazil from 2010 to 2023.
- Two independent white‑noise series (size 2^17 = 131072) synthetic public
Source sentence
3.2 White noise We have applied all the algorithms (the ones in the existing literature, and the ones proposed in this work) on two independent white noise series of size 217=131072 with the results presented in Fig. 3.
- Daily sugar and ethanol price series in Brazil market-prices public
Source sentence
We test different algorithms on synthetic series with an exact analytical solution, as well as on daily price series of ethanol and sugar in Brazil from 2010 to 2023. We substantiate these proposals through a comprehensive assessment of a synthetic dataset used in [11], uncorrelated white noise series, and daily sugar and ethanol price data in Brazil from 2010 to 2023.
Costs
Not applicable
Finds 5
- All newly proposed algorithms produce wider and more asymmetric multifractal spectra than the ABS (modulus MF‑DXA) approach when applied to the Brazil sugar‑ethanol price data. statistical
Source sentence
On the other hand, all the algorithms proposed in this work yield wider, more asymmetric spectra in comparison to the ABS (modulus MF_DXA) approach.
- The percentage of point pairs captured by the new algorithms varies from 20.4% (PM) to 67.5% (PS), with consistent multifractal parameters across options. statistical
Source sentence
The percentage of point pairs captured by different flavors of the algorithm varies between 20.4% (PM) and 67.5% (PS), and the resulting multifractal parameters are rather consistent across the set of options, as can be seen in Tab. 2. MF-DXA ABS MFCCA PS 67.5% MS 32.5% PB 57.4% MB 42.6% PP 28.4% PM 20.4% MP 22.1% MM 29.0% Figure 4.
- On uncorrelated white‑noise series, the standard MF‑DXA algorithm yields no results, while MF‑CCA produces divergent H(q) near q≈0 and a scattered f(α) spectrum. negative
Source sentence
As covariance between the two series has an equal chance of assuming positive and negative values the MF-DXA algorithm produces no results at all, and the MFCCA algorithm yields (for roughly half of the segments where the q-th order fluctuation function happens to be positive) the 𝐻(𝑞) function that diverges close to 𝑞~0, and a cloud of points for the 𝑓(𝛼) spectrum.
- The MS algorithm captures 32.5% of point pairs but shows distorted scaling for large segment sizes, pushing h(q) downwards for positive q and shifting the f(α) spectrum leftward. statistical
Source sentence
For example, the MS algorithm captures 32.5% of point pairs, but the scaling is distorted for very large segment sizes, pushing the ℎ(𝑞) curve downwards for positive 𝑞 values, and thus the 𝑓(𝛼) spectrum to the left.
- The PP algorithm, with 28.4% point‑pair capture, provides the most linear log‑log plots across q values, indicating the most consistent scaling among detailed‑restriction methods. statistical
Source sentence
The most consistent of more detailed restriction algorithms (the most “linear” of the log-log plots for different 𝑞 values) appears to be PP with 28.4% point pairs captured.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 18 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (18)
- On the fractal properties of cosmic rays and Sun dynamics cross-correlations (2022) — link
- Multifractal Cross-Correlations of Bitcoin and Ether Trading Characteristics in the Post-COVID-19 Time (2022) — link
- Multifractal characterization and cross correlations of reference evapotranspiration time series of India (2021) — link
- Multifractal detrended cross-correlation analysis on benchmark cryptocurrencies and crude oil prices (2020) — link
- Multifractal detrended cross-correlation analysis of wind speed and solar radiation. (2020) — link
- MULTIFRACTAL ANALYSIS OF BRAZILIAN AGRICULTURAL MARKET (2020) — link
- Multifractal detrended cross-correlation analysis between respiratory diseases and haze in South Korea (2020) — link
- Multifractal Detrended Cross-Correlation Analysis of Global Methane and Temperature (2020) — link
and 10 more
- Detecting correlations and triangular arbitrage opportunities in the Forex by means of multifractal detrended cross-correlations analysis (2019) — link
- Fractals (2018) — link
- Cross-correlations between the US monetary policy, US dollar index and crude oil market (2017) — link
- Detrended cross-correlations between returns, volatility, trading activity, and volume traded for the stock market companies (2015) — link
- The Economics of Brazil’s Ethanol-Sugar Markets, Mandates, and Tax Exemptions (2015) — link
- Detecting and interpreting distortions in hierarchical organization of complex time series. (2015) — link
- Detrended cross-correlation analysis consistently extended to multifractality. (2013) — link
- Multifractal detrended cross-correlation analysis for two nonstationary signals. (2008) — link
- Multifractal Detrended Fluctuation Analysis of Nonstationary Time Series (2002) — link
- Plenum press, New York, NY, USA (1988)
0 of 18 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.