docs/examples.md line 44 says:
MicroDataFrame.cov() and .corr() retain pandas' unweighted calculations and return plain pandas DataFrame matrices.
The first clause has been false since #330 merged (2026-09-18): both are frequency-weighted, and each cell of the matrix is the estimator applied to that pair of columns. docs/api.md already says so. A reader who checks the docs against the code finds a direct contradiction.
Fix: rewrite the paragraph to describe the weighted behaviour, keep the accurate note that the return type is a plain DataFrame, and add a two-line example whose numbers match numpy.cov(fweights=...) on the replicated sample.
docs/examples.mdline 44 says:The first clause has been false since #330 merged (2026-09-18): both are frequency-weighted, and each cell of the matrix is the estimator applied to that pair of columns.
docs/api.mdalready says so. A reader who checks the docs against the code finds a direct contradiction.Fix: rewrite the paragraph to describe the weighted behaviour, keep the accurate note that the return type is a plain
DataFrame, and add a two-line example whose numbers matchnumpy.cov(fweights=...)on the replicated sample.