| Conceptual problem |
Explanation of the same aspect of the data twice: first as noise, and second as a new discovery non-independent of this noise. |
Explanation of the same aspect of the data twice: first as existing knowledge, and second as new discovery redundant with this knowledge. |
| Statistical problem |
Model overfitting that results in high variance (low precision) of estimated model parameters. |
Model overspecification that results in high bias (low trueness) of estimated model parameters. |
| Statistical solution |
Tests of independence against sampled data in which all benchmark (noise) features are maximally random and all other aspects of the data are preserved. |
Tests of non-redundancy against sampled data in which all benchmark (existing knowledge) features are preserved and all other aspects of the data are maximally random. |
