Observer 2
|
Observer 1
|
|
|
present or positive
|
absent or negative
|
subtotal
|
present or positive
|
a
|
b
|
a + b
|
absent or negative
|
c
|
d
|
c + d
|
subtotal
|
a + c
|
b + d
|
a + b + c + d
|
observed agreement as a proportion =
= (a + d) / (a + b + c + d)
expected agreement by chance as a proportion =
= (((a + b) * (a+c)) + ((c + d) * (b + d))) / ((a + b + c + d)^2)
observed agreement as a frequency (raw count) =
= (a + d)
expected agreement by chance as a frequency =
= (((a + b) * (a+c)) + ((c + d) * (b + d))) / (a + b + c + d)
The kappa coefficient can either be calculated using the proportion or fraction of the total number, or by using the raw count. The difference is that each proportion is multiplied by the total number, since the proportion = (number showing feature) / (total number).
kappa by proportion =
= ((observed agreement as a proportion) – (expected agreement by chance as a proportion)) / (1 – (expected agreement by chance as a proportion))
kappa by frequency =
= ((observed agreement as a frequency) – (expected agreement by chance as a frequency)) / ((a + b + c + d) – (expected agreement by chance as a frequency))
standard deviation =
= SQRT (((observed agreement) * (1 – (observed agreement))) / (((total number) * ((1 – (expected agreement by chance))^2)))
95% confidence interval =
= (calculated kappa) +/- (1.96 * (standard deviation))
Interpretation:
• minimum value for kappa statistic: < 0
• maximum value: 1
• The higher the number, the greater the level of agreement between the 2 observers.
Result for Kappa
|
Strength of Agreement
|
< 0.00
|
poor
|
0.00 – 0.20
|
slight
|
0.21 – 0.40
|
fair
|
0.41 – 0.60
|
moderate
|
0.61 – 0.80
|
substantial
|
0.81 – 1.00
|
almost perfect
|
from page 165, Landis and Koch (1977)