SURVEY METHODS 



Sampling and Estimation 



The sample used in this survey was a national multi- 

 stage probability sample with each household existing in the 

 United States at the time of the survey having the same chance of 

 inclusion. Hence, estimates from the survey are unbiased in the 

 sampling sense and the sampling tolerances of estimates are measur- 

 able. The stages of selection were these: 



1. One hundred primary sample areas (52 metropolitan 

 areas and I4.8 non-metropolitan counties) contain- 

 ing 189 counties were first selected with known 

 probability. 



2. Within the 100 primary sample areas, UUO inter- 

 viewing clusters were selected with calculated 

 probabilities. These clusters, consisting of 

 entire Census Enumeration Districts or combina- 

 tions of city blocks, average about 200 dwelling 

 units each. 



3. Within the kkO clusters, sample households were 

 selected by random procedures from the prelist- 

 ings of all dwellings in the cluster. The ratio 

 of selection was fixed separately for each cluster 

 so that all households in the United States would 

 have the same probability of inclusion. In each 

 sample household the person responsible for planning 

 the meals was specified for interview. A total of 

 2,5U3 interviews were conducted. 



Because of the wide geographic dispersion of the sample 

 households given by the use of 100 primary sample areas and of large 

 interviewing clusters within these areas, the sampling tolerances 

 for many of the characteristics measured are not much greater than 

 those that would be obtained from a sample of households selected 

 completely at random from a list of all households in the country. 

 Sampling errors for all national estimates from the survey are suf- 

 ficiently low that the user may have confidence that the results 

 would not differ materially from those of a complete Census using 

 the same questionnaire and interviewing procedures. A detailed 

 listing of the calculated sampling errors of national estimates for 

 virtually all questions in the survey is given in the Appendix. 



