first visit. The c all- back means that each individual in the 

 sample is given two chances to be found at home out of six 

 evenings covered by the nights- at- home weighting, instead of 

 only one chance where no call- backs are made. The amount of 

 weighting is reduced accordingly about one-half, which results 

 in a corresponding reduction in the contribution to tolerance 

 due to the weighting. This extension of the not-at-home weight- 

 ing and its mathematical basis are fully described in an article 

 in the July, 195U issue of the Journal of Marketing Jy 



Sampling Reliabi lity of the Results 



The usefulness of a survey depends largely on the degree 

 of confidence that may be placed in its findings. Therefore users 

 need the assurance which comes from mathematical computation of 

 the tolerance limits, and these calculated tolerances must be suf- 

 ficiently small for the results to be useable. 



Among the many factors which contribute to sampling 

 efficiency, perhaps the most important is that of widespread geo- 

 graphic dispersion, not only for the country as a whole, but with- 

 in each locality and cluster. In this sample, 100 localities 

 (metropolitan areas and non-metropolitan counties) were included. 

 Since the variation between localities can be a major contribu- 

 tion to tolerances, the comparatively large number of localities 

 substantially reduced the tolerances that would otherwise have 

 been obtained from a sample of the same size. Because each cluster 

 was prelisted, individual households were selected from the list 

 at random intervals which assured widespread coverage of the en- 

 tire area in the cluster. This is equivalent to including many 

 more clusters where consecutive households are selected. 



There is always some risk involved in accepting and using 

 estimates based on any sample. It is literally the purpose of 

 statistical research to reduce these risks ox "chance" error as 

 far as possible. One unique advantage of a known probability sample 

 is that the degree of risk involved in using the projections can 

 be precisely stated. This means that the chance that any findings 

 would be different from those of a complete census survey of all 

 households in the country, using the same questionnaire and inter- 

 viewing methods, is known precisely. 



2/ "A Plan to Account for 'Not-At-Homes 1 by Combining Weighting 



and Call-backs", by WiHard R » Simmons, The Journal of Marketing , 

 July, ±9$h. 



303 



