This is the standard error as a proportion of the total sample 

 and must be converted back to represent a proportion of the 

 subsample . 



.0262 X 216 (size of total sample) - 5.6592 



5.6592 T 87 (size of subsample) = .06505 (say .065) 



This means that chances are, two out of three, that if all 

 eating places in Denver had been interviewed with the sanne 

 techniques used in this survey, this proportion would have 

 fallen between .396+ .065 and .396 - .065, or between 33.1 

 percent and 46. 1 percent. 



Below are the standard errors of the proportion estimated for 

 the corresponding percentages in all ten cities: 



Standard Error of the Proportion 



Atlanta, Georgia 

 Chicago, Illinois 

 Cleveland, Ohio 

 Denver, Colorado 

 Houston, Texas 

 Los Angeles, California 

 New York, New York 

 Omaha, Nebraska 

 Portland, Oregon 

 Springfield, Massachusetts 



14 



