OpenEpi, Version of December 1, 2003

Test Results for TwoByTwo Module

 

 

Dataset #

Left Side

a,b,c,d

Compared with

Result

 

 

 

 

 

1

Disease

20,6,0,5

ExactBB

Exact limits and p-values OK

2

Disease

2,2,4,6

6,4,2,0

3,0,1,1

8,4,1,1

19,10,8,8

 

 

Martin, Austin,  Epidemiology 1991; 2; p. 361

CMLE odds ratio, Fisher and mid P results agree.

MH chi is 0.374 vs. Mantel-Haenszel Summary Chi Square in OpenEpi of 0.1399. 

3

Exposure

27,95,44,443

SAS proc FREQ

Agreement

4

Exposure

23,74,20,134

4,21,24,309

SAS

Agreement except that method of calculating 2-sided confidence limits from 1-sided seems to be different.

5

Exposure

451,39,1729,636

ATFRAC

Agreement for OR based etiologic fraction and CLs.

6

Disease

451,1729,39,636

Stat22

Agreement, but OpenEpi unable to calculate Fisher and midP

7

Disease

Same as 6

ExactMA

Same result as 6

8

Disease

39,451,636,1729

ExactMA

Calculations agree and OpenEpi can compute, but p values “mostly zeroes”.

9

Disease

Same as 6

Stats Direct

Same as 6

10

Disease

Same as 8

Stats Direct

Same as 8.  Prevented fraction not available from Stats Direct.

11

Disease

72,20,684,553

Walter, IJE 1978;7;2. Table 2

Good agreement

12

Disease

4,386,4,1250

Rothman, EpiSheet, Downloaded Nov 2003

Good agreement.  2-sided confidence intervals calculated the same way as OpenEpi (doubling the minimum of the one-sided exact), and agree.

13

Disease

3,9,104,1059

1,3,5,86

 

Rothman, EpiSheet, Downloaded Nov 2003

Rothman values labeled RR and mhRR are actually odds ratios, and agree with OpenEpi odds ratios. 

14

Disease

8,5,98,115

22,16,76,69

Rothman, EpiSheet, Downloaded Nov 2003

ok

15

Disease

32,2,52407,18790

104,12,43248,10673

206,28,28612,5710

186,28,12663,2585

102,31,5317,1462

Rothman, EpiSheet, Downloaded Nov 2003

Rothman’s RR and mhRR are really odds ratios.  RDs and other parameters agree.

16

Exposure

27,95,44,443

Epi Info 2002, Ver 2

Good agreement

17

Exposure

1,2,3,4

Epi Info, Ver 3

Good agreement

18

Exposure

22,6,3,16

20,5,0,2

 

Epi Info, Ver 3

Good agreement

19

Exposure

43,11,3,18

Epi Info, Ver 3

Good agreement

20

Exposure

27,4,3,10

16,7,0,8

Epi Info, Ver 3

Good agreement

21

Exposure

1,0,2,3

Epi Info, Ver 3

Agreement. 

22

Exposure

66,28,36,32

139,61,93,54

Sullivan, Unpublished chapter, with hand-calculated results, downloadable with OpenEpi. Click HELP in TwobyTwo.

Agreement.

 

By Andy Dean with a great deal of help from Ray Simons, who kindly sent printouts of comparison materials, and Kevin Sullivan, whose excellent unpublished chapter on 2 x 2 tables and Person Time statistics is the source of the equations and the test examples for many of the functions. Agreement is considered satisfactory if the first two significant figures agree, when a number, such as a p-value, is small.

 

Conclusions: There are some issues to resolve about terminology for upper and lower one-sided exact confidence limits, and concerning whether to add 0.5 to each cell if a cell is 0, but otherwise TwoByTwo seems ready to release for serious use, although it is always a good idea to check important conclusions with more than one program.