CATRENDTEST procedure

Calculates the Cochran-Armitage chi-square test for trend (A.I. Glaser).

Option

PRINT = string token	Output required (test); default test

Parameters

DATA = tables	Table containing observed data
TREND = factors	Dimension of the table representing the trend; can default if only one dimension of size greater than 2
CHISQUARE = scalars	Saves the chi-square for trend
PROBABILITY = scalars	Saves the probability value for trend
DEVCHISQUARE = scalars	Saves the chi-square for deviations from a linear trend
DEVDF = scalars	Saves the degrees of freedom for the chi-square for deviations
DEVPROBABILITY = scalars	Saves the probability value for the chi-square for deviations

Description

The CATRENDTEST procedure calculates the Cochran-Armitage chi-square test for trend. Categorical data can be collected and categorized by explanatory factors (such as dosage or treatment level), and any analysis will try to indicate relationships between the response (binary) factor and explanatory factors. The Cochran-Armitage chi-square test calculates a chi-square statistic on 1 degree of freedom for a linear trend in the responses. The data are represented by a (2×K or K×2) contingency table, where K represents the explanatory factor (known as the trend).

The DATA parameter supplies the data values in a two-way table. The TREND parameter can be set to a factor to indicate which dimension of the table represents the trend; if this is omitted CATRENDTEST assumes that the trend is in the dimension with more than 2 rows or columns (the other dimension must have exactly 2 rows or columns).

By default CATRENDTEST prints the results of tests for trend and for deviation from a trend (chi-square values, degrees of freedom and probabilities), but you can suppress these by setting option PRINT=*.

Parameters CHISQUARE, PROBABILITY, DEVCHISQUARE, DEVDF and DEVPROBABILITY allow the results to be saved (in scalars).

Option: PRINT.

Parameters: DATA, TREND, CHISQUARE, PROBABILITY, DEVCHISQUARE, DEVDF, DEVPROBABILITY.

Method

The method is described in Section 15.2 of Armitage, Berry & Matthews (1994).

Reference

Armitage, P., Berry, G. & Matthews, J.N.S. (1994). Statistical Methods in Medical Research. Blackwell Science, Oxford.

See also

Procedures: MCNEMAR, QCOCHRAN.

Commands for: Basic and nonparametric statistics, Regression analysis.

Example

CAPTION     'CATRENDTEST example',\
            'Data from Table 15.1 of Armitage, Berry & Matthews (1994)';\
            STYLE=meta,plain
FACTOR      [LEVELS=2; LABELS=!T('Yes','No')] Attendence
FACTOR      [LEVELS=!(0,1,2,3)] Score
TABLE       [CLASS=Score,Attendence; VALUES=59,97,10,31,12,36,5,28] Patient
CATRENDTEST Patient; TREND=Score