Two Independent Proportions
Menu location: Analysis_Proportions_Two Independent.
This function examines the difference between two independent binomial proportions.
Another way of looking at two proportions is to put the counts/frequencies into a 2 by 2 contingency table and examine the relationship between the grouping into rows and the grouping into columns (see Fisher's exact test and 2 by 2 chi-square test).
| feature present | feature absent | |
| outcome positive: | a | b |
| outcome negative: | c | d |
r1 = a
r2 = b
n1 = a+c
n2 = b+d
p1 = r1/n1
p2 = r2/n2
proportion difference (delta) = p1-p2
For example, the proportion of students passing a test when taught using method A could be compared with the proportion passing when taught using method B.
StatsDirect provides an hypothesis test for the equality of the two proportions (i.e. delta = 0) and a confidence interval for the difference between the proportions. An exact two sided P value is calculated for the hypothesis test (null hypothesis that there is no difference between the two proportions) using a mid-P approach to Fisher's exact test. The conventional normal approximation is also given for the hypothesis test, you should only use this if the numbers are large and the exact (mid) P is not shown (Armitage and Berry, 1994).
Assumptions:
- two mutually exclusive outcomes
- two random samples
- samples from two independent populations
Technical validation
The iterative method of Miettinen and Nurminen is used to construct the confidence interval for the difference between the proportions (Mee, 1984; Anbar, 1983; Gart and Nam, 1990; Miettinen and Nurminen, 1985; Newcombe, 1998b). This "near exact" confidence interval will be in close but not in exact agreement with the exact two sided (mid) P value; i.e. just excluding zero just above P = 0.05.
Example
From Armitage and Berry (1994).
Two methods of treatment, A and B, for a particular disease were investigated. Out of 257 patients treated with method A 41 died and out of 244 patients treated with method B 64 died. We want to compare these fatality rates.
To analyse these data in StatsDirect you must select unpaired proportions from the proportions section of the analysis menu. Enter total observations in sample 1 as 257, number responding in sample 1 as 41, total observations in sample 2 as 244 and number responding in sample 2 as 64.
For this example:
Total 1 = 257, response 1 = 41
Proportion 1 = 0.159533
Total 2 = 244, response 2 = 64
Proportion 2 = 0.262295
Proportion difference = -0.102762
Approximate (Miettinen) 95% confidence interval = -0.17432 to -0.031588
Exact two sided (mid) P = .0044
Standard error of proportion difference = 0.03638
Standard normal deviate (z) = -2.824689
Approximate two sided P = .0047
Approximate one sided P = .0024
Here we can conclude that the difference between these two proportions is statistically significantly different from zero. With 95% confidence we can state that the true population fatality rate with treatment B is between 0.03 and 0.17 greater than with treatment A.