One cell count in a contingency table is unknown. Can you mask that cell or replace it with zero and run an ordinary…
StatisticsLevel: IntermediateBinary and categorical inference
One cell count in a contingency table is unknown. Can you mask that cell or replace it with zero and run an ordinary chi-square independence test?
Hints
- Distinguish an observed zero from an unobserved count.
- Check which row, column and grand totals would change.
Answer
Neither preserves the intended table: zero asserts no events, while an unknown count prevents determining its margins and expected frequencies. SciPy does not support masked cells for this test; recover the count or explicitly model the missing-data process before choosing an analysis.
Source and licence: SciPy — contingency tables · BSD-3-Clause · answer: Quantoros editorial