Transportation Statistics and Microsimulation

154    ◾    Transportation Statistics and Microsimulation

error sum of squares. An alternative is to use a rank test in which all

the data are placed in one column and ranked. e smallest data point

gets rank 1, the next smallest rank 2, …, and the largest is ranked as

n = n

1

+ n

2

+ … + n

g

. en the raw data are replaced by the ranked data

and ANOVA is performed on the latter. is is known as the Kruskal–

Wallis test, and it is very resilient to the presence of outliers. e descrip-

tion here is not the exact Kruskal–Wallis test, but a slight variant. Please

refer to Kruskal and Wallis (1952) and Ott and Longnecker (2009) for fur-

ther details. JMP ...