An order-directed score test for trend in ordered 2 X K tables

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Trend detection in ordered contingency tables is crucial for analyzing categorical data.
• Existing linear rank tests are sensitive to the choice of scoring systems, leading to variable p-values.
• A need exists for a robust trend detection method that is independent of pre-assigned scores.

Purpose of the Study:

• To present a new, score-free test procedure for detecting trends in ordered 2 x K contingency tables.
• To address the limitation of score-dependent p-values in existing linear rank tests.
• To provide a method easily generalizable for stratified analyses.

Main Methods:

• Development of an order-directed score statistic for trend detection.
• Application of the statistic to ordered 2 x K tables.
• Demonstration of generalization to stratified analysis involving multiple 2 x K tables.

Main Results:

• The proposed procedure effectively detects trends without requiring pre-assigned scores for ordinal categories.
• It circumvents the issue of varying p-values associated with different scoring systems.
• The method is shown to be adaptable for stratified data analysis.

Conclusions:

• The new test procedure offers a reliable and flexible approach for trend detection in ordered 2 x K tables.
• Its score-free nature enhances objectivity and consistency in statistical analysis.
• The generalizability to stratified analysis expands its utility in various research settings.