Example 10.9 Refer to the data set on mergency service of Example 10.5
(Table 10.2), but this time we investigate only one covariate, the workload
(hours). After fitting the second-degree polinomial model,
EðYiÞ¼siexpðb 0 þb 1 houriþb 2 hour^2 iÞwe obtained a result which indicates that thecurvature e¤ect is negligible
ðp¼ 0 : 8797 Þ.
Note:An SAS program would include the instruction
MODEL CASES = HOURS HOURS*HOURS/ DIST = POISSON
LINK = LOG OFFSET = LN;
The following is another interesting example comparing the incidences of
nonmelanoma skin cancer among women from two major metropolitan areas,
one in the south and one in the north.
Example 10.10 In this example, the dependent variable is the number of cases
of skin cancer. Data were obtained from two metropolitan areas: Minneapolis–
St. Paul and Dallas–Ft. Worth. The population of each area is divided into
eight age groups and the data are shown in Table 10.5.
TABLE 10.4
Variable Coe‰cient Standard Error zStatistic pValue
Intercept 8.1338 0.9220 8.822 <0.0001
No residency 0.2090 0.2012 1.039 0.2988
Female 0.1954 0.2182 0.896 0.3703
Revenue 0.0016 0.0028 0.571 0.5775
Hours 0.0007 0.0004 1.750 0.0452
TABLE 10.5
Minneapolis–St. Paul Dallas–Ft. WorthAge Group Cases Population Cases Population
15–24 1 172,675 4 181,343
25–34 16 123,065 38 146,207
35–44 30 96,216 119 121,374
45–54 71 92,051 221 111,353
55–64 102 72,159 259 83,004
65–74 130 54,722 310 55,932
75–84 133 32,185 226 29,007
85 þ 40 8,328 65 7,538
364 METHODS FOR COUNT DATA