Skip to main content icon/video/no-internet

In a randomized complete block design, it is often of interest to examine a set of c binary responses pertaining to the levels of some treatment condition provided either by a sample of participants used as their own controls to make these assessments or by a sample of c matched participants randomly assigned to each treatment level as members of a block. In 1950, the statistician William G. Cochran developed a test for the differences in the proportions of “success” among the treatment levels in c related groups in which the repeated measurements provided by each participant under c different conditions or the measurements provided by the c homogeneous participants in each block are binary responses, with success coded 1 and “failure” coded 0.

The Cochran Q test is a dichotomous data counterpart of another nonparametric procedure, the Friedman rank test, in which the responses to the c treatment levels within each block are ranked. Both these procedures are competitors of the classic two-way ANOVA randomized block F test, in which the responses to each treatment are measured on an interval or ratio scale and the assumption of underlying normality of the different treatment groups is made.

Motivation

Cochran's Q test has enjoyed widespread use in behavioral, business, educational, medical, and social science research when it may be desirable to evaluate possible significance of differences in the proportion of successes over several treatment groups.

As an example of a study in which each participant offers repeated measurements, one each for the c levels of the treatment or condition under evaluation, the efficacy of c different drugs prescribed for providing relief from some chronic condition may be investigated. Other examples of this type could involve taste testing wines or other consumer preference studies in which each of c items is classified as “acceptable” or “not acceptable.” As an example of a randomized complete block experiment in which the c homogeneous members of a block are randomly assigned, one each, to the c levels of the treatment, an educator may wish to form sets of homogeneous blocks of students and randomly assign the members of the block to each of c learning methods with the goal of assessing whether there are significant differences among the learning methods based on the proportions of successes observed.

None

Table 1 Data Layout for the Cochran Q Test

Development

The layout for the dichotomous responses from a sample of either r participants or r blocks of matched participants over c levels of a treatment condition is shown in Table 1.

Cochran's Q test statistic is given by

None

where

c is the number of treatment groups (i.e., columns),

r is the number of blocks (i.e., rows of subjects),

xij is the binary response (success = 1, failure = 0) for the jth treatment in the ith block,

x.j is the total number of successes for treatment j,

xi. is the total number of successes for block i, and

None = the total number of successes.

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Sage Recommends

We found other relevant content for you on other Sage platforms.

Loading