data taken from table in...

Dewailly E.  Poirier C.  Meyer FM.
Health hazards associated with windsurfing on polluted water.

American Journal of Public Health.  76(6):690-1, 1986 Jun.

Abstract

We documented the risks associated with windsurfing on sewage polluted
water. Seventy-nine windsurfers and 41 controls were studied over a
nine-day period for occurrence of symptoms of gastroenteritis, otitis,
conjunctivitis, and skin infection. Relative risks were 2.9 for occurrence
of one or more of these symptoms and 5.5 for symptoms of gastroenteritis.
Relative risk increased with the reported number of falls into the water.

dataset (revised June 18, 1997)

Variables

falls              0 = controls 
                   5 = those who fell  0-10 times
                  15 = those who fell 10-20 times
                  25 = those who fell 20-30 times
                  40 = those who fell  > 30 times
                  
sick               0 = no
                   1 = yes
                   
number             number of such individuals

denom              denominator ie #(sick = 0) + #(sick =1)

--------

NOTES

if you are fitting binary regression models, and if the program
(e.g. INSIGHT in SAS) allows you to specify a numerator and a denominator
as part of the model, then you could omit the observations with say sick = 0
( redundant since #(sick = 0) + #(sick =1) = denom  )
In INSIGHT, you specify the denominator variable in the box entitled "Binomial". 

Proc LOGISTIC in SAS  allows two formats to specify the model

  (i) MODEL variable = independents  
 (ii) MODEL events/trials = independents 
 
Format (i)
----------
 The response variable Y is a 0 or a 1 (it is assumed that the denominator
 is 1 unless explicitly told with a FREQ statement telling it to find the 'duplication
 factor' i.e. how many subjects have that value of Y. With (i), the Y=0 observations
 are SEPARATE from the Y=1 observations).

 PROC LOGISTIC;
  MODEL sick = falls;
  FREQ  number;

 You could also use format (i) in INSIGHT if the observations are individuals or
 if you have the sick and the not sick on different lines.   

Format (ii)
-----------
 Data a; INPUT falls sick number denom;
  IF sick = 1;
  cases = number;
  
 PROC LOGISTIC;
  MODEL cases/denom = falls;