Exact inference for categorical data: recent advances and continuing controversies

Area of Science:

• Statistics
• Biostatistics
• Data Analysis

Background:

• Exact statistical methods for small-sample categorical data have advanced significantly.
• Current methods often condition on sufficient statistics to eliminate unknown parameters.
• Algorithmic developments have improved the implementation of these exact methods.

Purpose of the Study:

• To summarize existing exact statistical approaches for categorical data.
• To describe recent advancements in exact methods.
• To examine the conservatism issue in exact methods due to data discreteness.

Main Methods:

• Review of exact statistical methods for categorical data.
• Analysis of interval estimation for proportions.
• Evaluation of exact methods for odds ratio calculation.
• Examination of the mid-P-value adjustment.

Main Results:

• Exact methods for small-sample categorical data are well-developed.
• Discreteness in data can lead to conservative results with some exact methods.
• Adjusted exact methods based on the mid-P-value appear effective in mitigating conservatism.

Conclusions:

• The mid-P-value adjustment offers a practical approach to address the conservatism of exact methods.
• Further research into exact methods is warranted, particularly for interval estimation and odds ratios.
• Adjusted exact methods provide a more balanced approach for small-sample categorical data analysis.