Standard Errors of the Mean or Total 



The standard errors of the quantities shown in certain tables 

 (as for instance, total quantity purchased of a particular item 

 of fish or shellfish) are naturally larger than the standard 

 errors of the proportions. For this reason, and because res- 

 ponse error may be high--when respondents recollect quan- 

 tities of items purchased--standard errors of the quantities 

 have not been computed. 



It should be noted, also, that in computing standard errors of 

 the quantities, the adjustment factors listed above cannot be 

 applied. It is necessary to ascertain adjustment factors for 

 each individual fish or shellfish itenn purchased, inasmuch as 

 the number of establishments purchasing each item will vary 

 with the individual case. 



B, Nonresponse Error 



Nonresponse error occurs when an establishment falls in a 

 probability sample, but no interview can be completed at the 

 establishment. To the extent that nonrespondents are different 

 from respondents the survey results might have been altered 

 had the nonrespondents been included in the interviewed sample, 



The completion rate for each city is given in Section II-C, 

 Sample Design, It varies from 80 percent in Cleveland to 96 

 percent in Atlanta. These completion rates are relatively high 

 indicating that nonresponse error has been held to a mininnum. 



As a standard part of field procedure, repeated call backs 

 were made to those establishments when no interview was com- 

 pleted at the time of the first visit. These procedures substan- 

 tially reduced the number of noncompleted interviews, and 

 lowered the possibility of nonresponse error. 



Where no interview was completed after repeated call backs, 

 several factors were operating: 



1. Absence of the proprietor or a qualified represen- 

 tative, during the period of the field work. 



15 



