=================================================================
   COMPARE2 Version 1.90
   Tuesday, 13th January 2009, 15:16.
   -----------------------------------------------------------------
   Measures with number-of-individuals denominators [module B]
   =================================================================

   DATA:
   A: Yes, 751  No, 7755  Denominator, 8506
   B: Yes, 623  No, 7479  Denominator, 8102

   RESULTS:
   Proportions (of "Yes"):  A, 0.0883  B, 0.0769
                                If inverse sampling was used,
                                see results at end of output.

   [Campbell (2007) recommends use of Upton's chi-square
    (the "N-1" chi-square) for this table, where no cell has
    an expected value less than 1.]

   Exact probabilities not computed.

   Upton's "N - 1" chi-sq.   =  7.101  P = 0.008 [ 7.7E-3 ]
   Pearson's chi-square      =  7.101  P = 0.008 [ 7.7E-3 ]
     with Yates's correction =  6.952  P = 0.008 [ 8.4E-3 ]
   Haber's adjusted chi-sq.  =  7.015  P = 0.008 [ 8.1E-3 ]

   Log-likelihood chi-square =  7.114  P = 0.008 [ 7.6E-3 ]
     with Yates's correction =  6.964  P = 0.008 [ 8.3E-3 ]

   DIFFERENCE [A minus B] = 0.011  S.E. = 0.004
     Large-sample method (Fleiss), continuity-corrected:
       90% C.I. = 0.004 to 0.019
       95% C.I. = 0.003 to 0.020
       99% C.I. = 0.000 to 0.023
     Wilson's score method:
      Not continuity-corrected (Newcombe's method 10):
       90% C.I.  = 0.004 to 0.018
       95% C.I.  = 0.003 to 0.020
       99% C.I.  = 0.000 to 0.022
      Continuity-corrected (Newcombe's method 11):
       90% C.I.  = 0.004 to 0.019
       95% C.I.  = 0.003 to 0.020
       99% C.I.  = 0.000 to 0.022

   RATIO [A:B] = 1.148     S.E. of log ratio = 0.052
     Traditional (log-transformation) method:
       90% C. I. = 1.05 to 1.25
       95% C. I. = 1.04 to 1.27
       99% C. I. = 1.00 to 1.31
     Zou-Donner method:
       90% C.I.  = 1.05 to 1.25
       95% C.I.  = 1.04 to 1.27
       99% C.I.  = 1.00 to 1.31
     Low-bias estimator of ratio in population = 1.147

   ODDS RATIO  = 1.16  [reciprocal = 0.86]
     Fisher's exact confidence intervals:
       90%:  1.06 to 1.28
       95%:  1.04 to 1.30
       99%:  1.00 to 1.35
     Mid-P exact confidence intervals:
       90%:  1.06 to 1.28
       95%:  1.04 to 1.30
       99%:  1.01 to 1.35
     Cornfield's confidence intervals:
       90%:  1.06 to 1.28
       95%:  1.04 to 1.30
       99%:  1.00 to 1.35

     S.E. of log O.R. = 0.057
     Low-bias indicator of O.R. in the population = 1.16
     Adjusted O.R. (0.5 added in each cell) = 1.16

   Yule's Q = 0.08        Phi = 0.02

   Lambda (for prediction of Yes-No distribution) = 0.00

   NUMBER NEEDED:
   Subjects needed in B to avoid 1 event*: 87.7
     95% C.I.: 50.4 to 341
   Assuming a risk of R per 1 in group B, number needed
     in group B to avoid 1 event* (if the risk ratio seen
     in this study sample is appropriate) = 7/R
     [Formula not valid if R exceeds 0.9 per 1]
   *If this is a therapeutic trial where A = "treated" and
    B = "controls", this is the NNTH (number needed to
    treat to produce one episode of harm).

   MEASURES OF IMPACT
     (if Group A is exposed, and Group B not exposed, to
     a risk factor]
    Attributable fraction in exposed = 12.9%
     90% C.I. = 5.1 to 20.0
     95% C.I. = 3.6 to 21.3
     99% C.I. = 0.4 to 23.8
    Attributable fraction in population = 7.1%
     90% C.I. = 2.7% to 11.4%
     95% C.I. = 1.9% to 12.2%
     99% C.I. = 0.2% to 13.8%

   INVERSE SAMPLING
    The following results are applicable if, in each group
    (A and B), subjects were added until a prespecified
    number of cases (subjects with "Yes") were found.

       Large-sample test: one-tailed P = 0.002 [ 2.1E-3 ]
                          two-tailed P = 0.004 [ 4.2E-3 ]
       DIFFERENCE [A minus B] = 0.011
                       S.E.   = 0.004
         Unbiased estimate    = 0.011
          Confidence intervals:
           90% conf. interval = 0.004 to 0.018
           95% conf. interval = 0.003 to 0.020
           99% conf. interval = 0.000 to 0.022
       RATIO (A:B)            = 1.148
         Unbiased estimate    = 1.147
          Alternative confidence intervals:
           90% conf. interval = 1.054 to 1.251
                            or  1.050 to 1.256*
           95% conf. interval = 1.037 to 1.271
                            or  1.033 to 1.277*
           99% conf. interval = 1.004 to 1.313
                            or  0.999 to 1.321*
          * May be preferable if the numbers or proportions
            of cases (subjects with "Yes") are small.
       ODDS RATIO   = 1.16  [reciprocal = 0.86]
           90% C.I. = 1.06 to 1.28
           95% C.I. = 1.04 to 1.30
           99% C.I. = 1.00 to 1.34


   ======================================================*===========