KruskalWallis

1
Kruskal-Wallis

• H test

2
Kruskal-Wallis

• Hypotheses
• Ho the k distributions are identical
• H1 not all distributions are the same
• Or
• Ho there is no difference among the k groups
  with respect to y
• H1 at least one of the k groups of x differs
  from the other with respect to y

3
Kruskal-Wallis

• Critical Value
• a
• d. f. k 1
• Where k number of groups being compared
• The statistic follows the chi-squared
  distribution.

4
Kruskal-Wallis

• Test Statistic

5
Kruskal-Wallis

• Where nT sample size
• ni number of observations in group i
• Ti the sum of the ranks for group i

6
Kruskal-Wallis

• Example
• The table below contains the ratings given to
  three new television programmes (the distribution
  of these ratings are too skewed). Is there a
  significant difference among the groups?