The interpretation above of the concept of an odds ratio as an odds can be
generalized as follows. The aim here is to present an e‰cient method for use
with ordered 2kcontingency tables, tables with two rows andkcolumns
having a certain natural ordering. The figure summarized is the generalized
odds formulated from the concept of odds ratio. Let us first consider an exam-
ple concerning the use of seat belts in automobiles. Each accident in this
example is classified according to whether a seat belt was used and to the
severity of injuries received: none, minor, major, or death (Table 1.13).
To compare the extent of injury from those who used seat belts with those
who did not, we can calculate the percent of seat belt users in each injury group
that decreases from level ‘‘none’’ to level ‘‘death,’’ and the results are:
None:
75
75 þ 65
¼54%
Minor:
160
160 þ 175
¼48%
Major:
100
100 þ 135
¼43%
Death:
15
15 þ 25
¼38%
What we are seeing here is atrendor anassociationindicating that the lower
the percentage of seat belt users, the more severe the injury.
We now present the concept ofgeneralized odds, a special statistic specifi-
cally formulated to measure the strength of such a trend and will use the same
example and another one to illustrate its use. In general, consider an ordered
2 ktable with the frequencies shown in Table 1.14.
TABLE 1.13
Extent of Injury Received
Seat Belt None Minor Major Death
Yes 75 160 100 15
No 65 175 135 25
TABLE 1.14
Column Level
Row 1 2 k Total
1 a 1 a 2 ak A
2 b 1 b 2 bk B
Total n 1 n 2 nk N
RATIOS 23