Compute Scores for Ordinal Contingency Tables

A utility function computing score transformations of raw data, including normal scores, exponential scores, and Savage scores, typically used as a preprocessing step for nonparametric tests.

Usage

scores(x, MARGIN = 1, method = c("table", "ranks", "ridit", "modridit"))

Arguments

x

a contingency table (matrix or array of counts).

MARGIN

an integer indicating the margin over which to compute the scores. Defaults to 1 (rows). Use 2 for columns.

method

a character string specifying the scoring method. One of:

"table": uses numeric dimnames if available, otherwise assigns sequential integers.
"ranks": mid-ranks based on cumulative frequencies.
"ridit": ridit scores (ranks divided by total count).
"modridit": modified ridit scores (ranks divided by total count + 1).

Value

A numeric vector of scores corresponding to the levels of the selected margin.

Details

Computes score values for the levels of a contingency table margin. These scores are used in several statistical procedures such as the Cochran-Armitage test and correlation measures for ordinal data.

The function supports different scoring methods, including simple table-based scores, ranks, and ridit-type transformations.

For method = "table", numeric dimension names are used as scores if available. Otherwise, consecutive integers starting from 1 are assigned.

For rank-based methods, scores are computed as midpoints of cumulative frequencies along the selected margin.

Ridit and modified ridit scores are normalized versions of these ranks.

References

Lecoutre, E. (2005). R-help mailing list discussion. https://stat.ethz.ch/pipermail/r-help/2005-July/076371.html