“A new test of independence” does not identify one universal statistical procedure. Several papers use similar titles for methods designed for different data: 2×2 categorical tables, paired bivariate observations, metric-space data such as time series, or high-dimensional variables. Choose a test for the structure of your data and the kinds of dependence you need it to detect—not simply because a method is described as new.
What does an independence test ask?
For two variables or random elements, independence means their joint behavior can be explained by the product of their separate marginal behaviors. In practical terms, knowing one variable does not provide information about the other. An independence test assesses whether the observed data are consistent with that relationship.
The title alone is ambiguous: multiple scholarly papers add qualifiers such as “for bivariate observations,” “in 2×2 contingency tables,” “based on recurrence rates,” or “for high-dimensional data.” These are distinct proposals, not editions of one test. Their results and intended use cannot be treated as interchangeable.
Which method matches your data?
| Data structure | Method and source | Core idea or comparison | Evidence and calibration described |
|---|---|---|---|
| Two categorical variables in a 2×2 table | Piotr Sulewski, “A New Test for Independence in 2×2 Contingency Tables” (2017) | Compares the common chi-square test, a modular test, a d-square modification of Pearson’s test, and a proposed logarithmic-minimum test. | Critical values are obtained by Monte Carlo methods; the paper compares power. Specific sample sizes and power values are not stated in the available record (Sulewski, 2017). |
| Paired observations on two variables | Dimitrios Bagkavos and Prakash N. Patil, “A new test of independence for bivariate observations” (2017) | Uses the fact that, under independence, every conditional quantile of one variable given the other is constant. | The paper discusses asymptotic distributions under the null and alternative, an Edgeworth expansion, a bandwidth-selection rule, and numerical comparisons. Specific calibration values and computational costs are not stated in the available record (Bagkavos and Patil, 2017). |
| Random elements in metric spaces, including random variables, vectors, or time series | Juan Kalemkerian and Diego Fernández, “An Independence Test Based on Recurrence Rates” (posted to arXiv on 9 August 2019) | Defines a Cramér–von Mises-type functional applied to a U-process built from recurrence rates. It uses distances and information across all possible recurrence-radius values rather than selecting one pair of thresholds. | The method is described for metric-space data; specific finite-sample calibration values and computational costs are not stated in the available record (Kalemkerian and Fernández, 2019). |
| High-dimensional data, where many variables are involved | Guangyu Mao, “A new test of independence for high-dimensional data” (2014) | Proposes a statistic for testing independence in a high-dimensional setting. | The record reports simulation performance comparable to existing tests, with higher power in some circumstances. The particular circumstances and numerical values are not stated in the available record (Mao, 2014). |
How to choose an independence test
- Identify the data form. Decide whether your observations form a 2×2 categorical table, paired bivariate measurements, vectors or time series with a meaningful distance, or a high-dimensional dataset. Start with methods designed for that structure.
- Specify the dependence you care about. Ask what kinds of departures from independence matter for your application. A method’s reported power applies to the alternatives and conditions studied; it does not establish that the method is best for every form of dependence.
- Check how the test is calibrated. Find out whether its reference distribution is justified asymptotically or obtained through a procedure such as Monte Carlo simulation, and whether that calibration suits your sample and design. The methods listed here do not all have the same calibration evidence.
- Review the evidence for your setting. Look for theoretical results, simulations under relevant alternatives, and real-data applications. A simulation comparison is evidence about the conditions simulated, not a universal ranking.
- Assess practical constraints. Check the method’s assumptions, tuning choices such as bandwidths or thresholds, and computational demands in the original paper before applying it. The available summaries do not establish a single cross-method comparison of computational cost.
What these newer tests do—and do not—show
The proposals address different obstacles: the 2×2 paper compares several statistics for a specific table structure; the bivariate method uses conditional quantiles; the recurrence-rate method is formulated through distances so it can extend to metric-space objects; and Mao’s method targets high-dimensional data. The recurrence-rate authors say their construction uses information across possible radius values rather than choosing one pair of recurrence thresholds.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11That variety is also why “new” is not a verdict. The records describe theory, Monte Carlo or simulation comparisons, and numerical evidence, but provide no common benchmark that ranks these methods against one another across data types. No single headline power figure or universal winner follows from the papers’ titles or summaries.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What to read before applying one
Use the complete paper that matches your data structure to verify its assumptions, implementation details, calibration, and studied alternatives. The publication details are: Sulewski’s 2017 article in Acta Universitatis Lodziensis. Folia Oeconomica; Bagkvos and Patil’s 2017 article in the Journal of Multivariate Analysis; Kalemkerian and Fernández’s arXiv paper posted 9 August 2019; and Mao’s 2014 article in Statistics & Probability Letters. The bibliographic records summarized here do not establish exact benchmark values or a shared comparison across all four approaches.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
