You are on page 1of 19

15 Distribution-Free

Procedures

Copyright © Cengage Learning. All rights reserved.


15.4 Distribution-Free ANOVA

Copyright © Cengage Learning. All rights reserved.


Distribution-Free ANOVA
The single-factor ANOVA model for comparing I population
or treatment means assumed that for i = 1, 2, . . . , I,
a random sample of size Ji was drawn from a normal
population with mean i and variance  2.

This can be written as

Xij = I + ij j = 1, . . . , Ji; i = 1, . . . , I (15.14)

where the ij’s are independent and normally distributed


with mean zero and variance  2. The next procedure for
testing equality of the i’s requires only that the ij’s have
the same continuous distribution. 3
The Kruskal-Wallis Test

4
The Kruskal-Wallis Test
Let N = Ji, the total number of observations in the data
set, and suppose we rank all N observations from 1
(the smallest Xij) to N (the largest Xij).

When
H0: 1 = 2 = · · · = 1
is true, the N observations all come from the same
distribution, in which case all possible assignments of the
ranks 1, 2, . . . , N to the I samples are equally likely and
we expect ranks to be intermingled in these samples.

5
The Kruskal-Wallis Test
If, however, H0 is false, then some samples will consist
mostly of observations having small ranks in the combined
sample, whereas others will consist mostly of observations
having large ranks.

More specifically, if Rij denotes the rank of Xij among the N


observations, and Ri and Ri denote, respectively, the total
and average of the ranks in the ith sample, then when H0 is
true,

6
The Kruskal-Wallis Test
The Kruskal-Wallis test statistic is a measure of the extent
to which the Ri’s deviate from their common expected
value (N + 1)/2, and H0 is rejected if the computed value of
the statistic indicates too great a discrepancy between
observed and expected rank averages.

7
The Kruskal-Wallis Test
Test Statistic

(15.15)

The second expression for K is the computational formula; it


involves the rank totals (Ri’s) rather than the averages and
requires only one subtraction.

If H0 is rejected when k  c, then c should be chosen so that


the test has level . 8
The Kruskal-Wallis Test
That is, c should be the upper-tail critical value of the
distribution of K when H0 is true.

Under H0, each possible assignment of the ranks to the I


samples is equally likely, so in theory all such assignments
can be enumerated, the value of K determined for each
one, and the null distribution obtained by counting the
number of times each value of K occurs.

Clearly, this computation is tedious, so even though there


are tables of the exact null distribution and critical values
for small values of the Ji’s, we will use the following
“large-sample” approximation.
9
The Kruskal-Wallis Test
Proposition
When H0 is true and either
I=3 Ji  6 (i = 1, 2, 3)
or

I>3 Ji  5 (i = 1, . . . , I )

then K has approximately a chi-squared distribution with


I – 1 df.

This implies that a test with approximate significance level


 rejects H0 if k 
10
Example 9
The accompanying observations (Table 15.6) on axial
stiffness index resulted from a study of metal-plate
connected trusses in which five different plate lengths—4 in.,
6 in., 8 in., 10 in., and 12 in.—were used (“Modeling Joints
Made with Light-Gauge Metal Connector Plates,” Forest
Products J., 1979: 39–44).

Data and Ranks for Example 9


Table 15.6 11
Example 9 cont’d

The computed value of K is

At level .01, = 13.277, and since 20.12  13.277, H0 is


rejected and we conclude that expected axial stiffness does
depend on plate length.

12
Friedman’s Test for a Randomized
Block Experiment

13
Friedman’s Test for a Randomized Block Experiment

Suppose
Xij =  + i + j + ij,

where i is the ith treatment effect, j is the jth block effect,


and the ij’s are drawn independently from the same
continuous (but not necessarily normal) distribution.

Then to test
H0: 1 = 2 = · · · = 1 = 0,

the null hypothesis of no treatment effects, the observations


are first ranked separately from 1 to I within each block, and
then the rank average ri is computed for each of the I
treatments. 14
Friedman’s Test for a Randomized Block Experiment

When H0 is true, the ri’s should be close to one another,


since within each block all I! assignments of ranks to
treatments are equally likely.

Friedman’s test statistic measures the discrepancy


between the expected value (I + 1)/2 of each rank
average and the ri’s.

15
Friedman’s Test for a Randomized Block Experiment

Test Statistic

As with the Kruskal-Wallis test, Friedman’s test rejects H0


when the computed value of the test statistic is too large.

For the cases I = 3, J = 2, . . . , 15 and I = 4, J = 2, . . . , 8,


Lehmann’s book gives the upper-tail critical values for the
test.

16
Friedman’s Test for a Randomized Block Experiment

Alternatively, for even moderate values of J, the test


statistic Fr has approximately a chi-squared distribution with
I – 1 df when H0 is true, so H0 can be rejected if fr 

17
Example 10
The article “Physiological Effects During Hypnotically
Requested Emotions” (Psychosomatic Med., 1963:
334–343) reports the following data (Table 15.7) on skin
potential (mV) when the emotions of fear, happiness,
depression, and calmness were requested from each of
eight subjects.

Data and Ranks for Example 10


Table 15.7
18
Example 10 cont’d

Thus

fr = (1686) – 3(8)(5)

= 6.45

At level .05, = 7.815, and because 6.45 < 7.815, H0 is


not rejected.

There is no evidence that average skin potential depends


on which emotion is requested.

19

You might also like