What does MIC stand for?
| Full form | Maximal Information Coefficient |
|---|---|
| Category | Science & engineering |
| Type | Initialism |
The initialism MIC stands for Maximal Information Coefficient in statistics. It is a statistic used to explore relationships between pairs of variables, including relationships that a straight-line summary can miss.
MIC uses a grid-based approach to summarize dependence between two variables. The authors’ published 2016 methodological paper distinguishes sample MIC, the population quantity MIC* and the later estimator MICe. Keep the exact method name when reporting a result: a column labelled MICe should not be relabelled as the original MIC without checking what was computed. Pairwise association also describes a different question from why a relationship occurs. For a concrete illustration, take the five artificial pairs (−2, 4), (−1, 1), (0, 0), (1, 1), (2, 4). Every y value equals x squared, yet the symmetric points have zero Pearson linear correlation. The pattern is U-shaped, so zero linear correlation does not mean that these supplied values lack a relationship. This illustration computes no MIC score and is not a benchmark favouring one statistic. In a real analysis, retain the variables, sample size, software, estimator and settings alongside a reported score. Then inspect the relationship in its substantive context. A score used to rank pairs for inspection is not, by itself, a demonstration that changing one variable causes the other to change. The practical lesson is to name the method accurately and state what the analysis actually examined.
Reshef et al.: Measuring Dependence Powerfully and Equitably, JMLR 2016 ↗
Meaning in context
- Maximal
- The statistic involves maximizing over grid-based summaries within its definition.
- Information Coefficient
- An association statistic; preserve its distinction from MIC* and MICe in a methods description.
Practical examples
Inspect a nonlinear pattern
- Before
Toy pairs: (−2,4), (−1,1), (0,0), (1,1), (2,4). Draft: Their Pearson correlation is zero, so the variables have no relationship.
- After
The toy pairs follow y = x² despite zero Pearson linear correlation.
The sum of x(y − 2) is zero by symmetry, while both variables vary. This explains the zero linear correlation without denying the visible U-shaped relationship. No MIC value is calculated here.
Name the estimator actually reported
- Before
Fictional methods record: estimator MICe, software build Cedar, 240 paired observations. Draft: We used the original MIC implementation.
- After
The fictional record reports MICe using build Cedar on 240 paired observations.
Preserve the estimator label, implementation and sample size. Related estimators are not interchangeable names for one identical computation.
Original illustrative examples, not quotations or reports of actual events.
Read a relationship and its method
What does a reported MIC result tell me, and which method was actually used?
See all guidance
Read a result table
Retain the two variables and exact estimator label.
The fictional MICe result concerns temperature and yield in the stated sample.
Describe a pattern
Inspect the paired values as well as the summary statistic.
These five toy points follow a U-shaped pattern.
Write the conclusion
State an association before considering any separate causal argument.
The analysis ranks the pair for further inspection.
Retain the estimator and analysis context; inspect the relationship rather than treating a linear summary or association score as a causal explanation.
Help improve this entry
Spot a mistake or have a clearer example? Suggest a change.
By AcronymIndex Editorial Team · Editorial policy