), where the observed response can take only one of two possible values; for example a treatment may result in a ‘success’ or ‘failure’. The data is recorded as either

1

0

to represent this dichotomization.

The use of this ‘randomized block design’ allows the effect of differences between the blocks to be separated from the differences between the treatments. The test assumes that the blocks were randomly selected from all possible blocks and that the result may be one of two possible outcomes common to all treatments within blocks.

The null and alternative hypotheses to be tested may be stated as follows.

$H_{0}$ :	the treatments are equally effective, that is the probability of obtaining a $1$ within a block is the same for each treatment.
$H_{1}$ :	there is a difference between the treatments, that is the probability of obtaining a $1$ is not the same for different treatments within blocks.

The data is often represented in the form of a table with the

n

rows representing the blocks and the

k

columns the treatments. Let

R_{i}

represent the row totals, for

i = 1, 2, \dots, n

, and

C_{j}

represent the column totals, for

j = 1, 2, \dots, k

. Let

x_{i j}

represent the response or result where

x_{i j} = 0

1

Blocks	Treatments				Row Totals
Blocks	1	2		$k$	Row Totals
1	$x_{11}$	$x_{12}$	$\dots$	$x_{1 k}$	$R_{1}$
2	$x_{21}$	$x_{22}$	$\dots$	$x_{2 k}$	$R_{2}$
$⋮$			$⋮$		$⋮$
$n$	$x_{n 1}$	$x_{n 2}$	$\dots$	$x_{n k}$	$R_{n}$
Column Totals	$C_{1}$	$C_{2}$		$C_{k}$	$N = Grand Total$

p_{i j} = \Pr (x_{i j} = 1)

, for

i = 1, 2, \dots, n

and

j = 1, 2, \dots, k

, then the hypotheses may be restated as follows

$H_{0}$ :	$p_{i 1} = p_{i 2} = \dots = p_{i k}$ , for each $i = 1, 2, \dots, n$ .
$H_{1}$ :	$p_{i j} \neq p_{i k}$ , for some $j$ and $k$ , and for some $i$ .

The test statistic is defined as

Q = \frac{k (k - 1) \sum_{j = 1}^{k} {(C_{j} - \frac{N}{k})}^{2}}{\sum_{i = 1}^{n} R_{i} (k - R_{i})} .

When the number of blocks,

n

, is large relative to the number of treatments,

k

Q

has an approximate

χ^{2}

-distribution with

k - 1

degrees of freedom. This is used to find the probability,

p

, of obtaining a statistic greater than or equal to the computed value of

Q

. Thus

p

is the upper tail probability associated with the computed value of

Q

, where the

χ^{2}

-distribution is used to approximate the true distribution of

Q

4 References

Conover W J (1980) Practical Nonparametric Statistics Wiley

Siegel S (1956) Non-parametric Statistics for the Behavioral Sciences McGraw–Hill

5 Arguments

1: $n$ – Integer Input: On entry: $n$ , the number of blocks.

Constraint: $n \geq 2$ .
2: $k$ – Integer Input: On entry: $k$ , the number of treatments.

Constraint: $k \geq 2$ .
3: $x (ldx, k)$ – Real (Kind=nag_wp) array Input: On entry: the matrix of observed zero-one data. $x (i, j)$ must contain the value $x_{i j}$ , for $i = 1, 2, \dots, n$ and $j = 1, 2, \dots, k$ .

Constraint: $x (i, j) = 0.0$ or $1.0$ , for $i = 1, 2, \dots, n$ and $j = 1, 2, \dots, k$ .
4: $ldx$ – Integer Input: On entry: the first dimension of the array x as declared in the (sub)program from which g08alf is called.

Constraint: $ldx \geq n$ .
5: $q$ – Real (Kind=nag_wp) Output: On exit: the value of the Cochran $Q$ -test statistic.
6: $prob$ – Real (Kind=nag_wp) Output: On exit: the upper tail probability, $p$ , associated with the Cochran $Q$ -test statistic, that is the probability of obtaining a value greater than or equal to the observed value (the output value of q).
7: $ifail$ – Integer Input/Output: On entry: ifail must be set to $0$ , $−1$ or $1$ to set behaviour on detection of an error; these values have no effect when no error is detected.
A value of $0$ causes the printing of an error message and program execution will be halted; otherwise program execution continues. A value of $−1$ means that an error message is printed while a value of $1$ means that it is not.

If halting is not appropriate, the value $−1$ or $1$ is recommended. If message printing is undesirable, then the value $1$ is recommended. Otherwise, the value $0$ is recommended. When the value $- 1$ or $1$ is used it is essential to test the value of ifail on exit.

On exit: $ifail = 0$ unless the routine detects an error or a warning has been flagged (see Section 6).

6 Error Indicators and Warnings

If on entry

ifail = 0

−1

, explanatory error messages are output on the current error message unit (as defined by x04aaf).

Errors or warnings detected by the routine:

$ifail = 1$: On entry, $k = ⟨ value ⟩$ .
Constraint: $k \geq 2$ .

On entry, $ldx = ⟨ value ⟩$ and $n = ⟨ value ⟩$ .
Constraint: $ldx \geq n$ .

On entry, $n = ⟨ value ⟩$ .
Constraint: $n \geq 2$ .

$ifail = 2$: On entry, $i = ⟨ value ⟩$ , $j = ⟨ value ⟩$ and $x (i, j) = ⟨ value ⟩$ .
Constraint: $x (i, j) = 0.0$ or $1.0$ .

$ifail = 3$: The solution has failed to converge while calculating the tail probability. The returned result may still be a reasonable approximation.

$ifail = - 99$: An unexpected error has been triggered by this routine. Please contact NAG.
See Section 7 in the Introduction to the NAG Library FL Interface for further information.

$ifail = - 399$: Your licence key may have expired or may not have been installed correctly.
See Section 8 in the Introduction to the NAG Library FL Interface for further information.

$ifail = - 999$: Dynamic memory allocation failed.
See Section 9 in the Introduction to the NAG Library FL Interface for further information.

7 Accuracy

The use of the

χ^{2}

-distribution as an approximation to the true distribution of the Cochran

Q

-test statistic improves as

k

increases and as

n

increases relative to

k

. This approximation should be a reasonable one when the total number of observations left, after omitting those rows containing all

0

1

, is greater than about

25

and the number of rows left is larger than

5

8 Parallelism and Performance

Background information to multithreading can be found in the Multithreading documentation.

g08alf is not threaded in any implementation.

9 Further Comments

None.

10 Example

The following example is taken from page 201 of Conover (1980). The data represents the success of three basketball enthusiasts in predicting the outcome of

12

collegiate basketball games, selected at random, using

1

for successful prediction of the outcome and

0

for unsuccessful prediction. This data is read in and the Cochran

Q

-test statistic and its corresponding upper tail probability are computed and printed.

g08al: FL CL CPP AD PY MB

NAG FL Interfaceg08alf (test_​cochranq)

▸▿ Contents

1 Purpose

2 Specification

3 Description

4 References

5 Arguments

6 Error Indicators and Warnings

7 Accuracy

8 Parallelism and Performance

9 Further Comments

10 Example

10.1 Program Text

10.2 Program Data

10.3 Program Results

NAG FL Interface
g08alf (test_cochranq)