Each reported survey figure will have its own tolerance 

 limits. These limits will depend on the distribution of the 

 characteristic under study among the various localities in the 

 country, among different neighborhoods within the same locality, 

 and among separate households within the same neighborhood (or 

 cluster) . The errors given in the table below' were calculated 

 by a procedure 2/ which takes into account the variations arising 

 from all stages of the sampling, together with contributions to 

 error of the nights-at-home weighting procedure, -urther, the 

 error calculated from the sample results by this method also in- 

 cludes the contribution to total error arising from interviewer 

 variability, but does not include any systematic bias which might 

 arise from non-sampling sources. 



TABLE OF SAMPLING ERRORS 

 FOR NATIONAL SAMPLE 



Ques. 

 No. 



C harac teri s tic s 



Percent 

 Having the 

 Character- 

 istic 



Standard 

 Error in 

 Percentage 

 Foints 



Characteristics in Section? A and B 

 Percent of all sample households 

 that: 



A-lb Served canned tuna in past twelve months 

 A-lb Served canned salmon in past twelve 



months 

 A-lb Served canned sardines in past twelve 



months 

 A-lb Served canned shrimp in past twelve 



months 

 A-lc Served canned tuna most frequently 

 A-lc Served canned salmon most frequently 

 A-lc Served canned sardines most frequently 

 A-lc Served canned shrimp most frequently 



76.1 

 68.8 

 50.3 



1.2 



1.2 



1.1 



2u.O 



1.6 



53.1 



2.0 



21 .U 



1.5 



$.h 



0.3 



1.5 



0.3 



(Continued) 





3y The procedure used depends on estimating the variance of the 

 whole sample from the variances between systematic sub-samples. 



30li 



