k¼
ð 2 Þ½ð 35 Þð 35 Þð 20 Þð 10 Þ
ð 45 Þð 45 Þþð 55 Þð 55 Þ
¼ 0 : 406
indicates that the agreement is barely acceptable.
It should also be pointed out that:
- The kappa statistic, as a measure for agreement, can also be used when
there are more than two categories for classification:
k¼
P
ipii
P
ipiþpþi
1
P
ipiþpþi
- We can form category-specific kappa statistics (e.g., with two categories);
we have
k 1 ¼
p 11 p 1 þpþ 1
1 p 1 þpþ 1
k 2 ¼
p 22 p 2 þpþ 2
1 p 2 þpþ 2
- The major problem with kappa is that it approaches zero (even with a
high degree of agreement) if the prevalence is near 0 or near 1.
3.2 NORMAL DISTRIBUTION
3.2.1 Shape of the Normal Curve
The histogram of Figure 2.3 is reproduced here as Figure 3.1 (for numerical
details, see Table 2.2). A close examination shows that in general, the relative
frequencies (or densities) are greatest in the vicinity of the intervals 20–29, 30–
39, and 40–49 and decrease as we go toward both extremes of the range of
measurements.
TABLE 3.6
Nurse 2
Nurse 1 Normal Not Normal Total
Normal 35 10 45
Not normal 20 35 55
Total 55 45 100
120 PROBABILITY AND PROBABILITY MODELS