Bartlett's test for homogeneity of variance A test, introduced by
Bartlett in 1937, for equality of
variance in
k populations having
normal distributions with unknown means. The
data consist of random
samples of sizes
n1,
n2,…,
nk, with the total number of
observations being denoted by
n and the
unbiased estimates of the population variances being . Denoting (n
j − 1) by
vj, for all
j, Bartlett's test statistic,
L, is defined bywhere
v =
v1 +
v2 + ··· +
vk. Unusually low values of
L indicate unequal variances. The test is affected by departures from normality. See also
Cochran's C-test,
Levene's test.
No comments:
Post a Comment