Dataset # |
Confidence, Power (%) |
Ratio, Controls with exposure, Cases with
exp. |
Compared with |
Result |
S-2 |
95,90 |
2, 30%, RR>=4 |
Schlesselman, JJ, Case-Control
Studies, |
Agreement (34 and 68) |
S-6a |
95,80 |
1,40%,60% |
Pezzullo JavaStat website. Proportion Difference Power/Sample Size Calculation http://members.aol.com/johnp71/proppowr.html |
Agreement (107 vs 108 due to rounding up in OpenEpi) |
S-6b |
95,80 |
1, 40%,60% |
Epi Info Statcalc |
Agreement (107 vs 108 due to rounding up in OpenEpi) |
S-3a |
95,80 |
4, 5%,10% |
Pezzullo JavaStat website. Proportion Difference Power/Sample Size Calculation http://members.aol.com/johnp71/proppowr.html |
273 vs 274 and 1094 vs 1097 Pezzullo requires entering 0.25 for ratio rather than 4, and the difference appears to be in which group is calculated first and then adjusted by the ratio to find the number for the other group. |
S-3b |
same |
same |
Epi Info Statcalc |
273 vs. 274, 1097 vs 1092 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Testing by Andy Dean. .
Conclusions: OpenEpi sample size programs round results UP rather than to the nearest integer, since it is difficult to find 0.1 person to participate in a study. Agreement takes this into account. Differences in S-3 should be investigated, but are not serious for practical work.
These results are sufficient to remove the BETA labels from the programs, although further testing and comparison of important results with those from other programs is always in order. Please send us results of further testing so that we can add them to this list.