Problems of quota definition for non-random surveys
The first part of this paper attempts to enumerate the problems which face the sampling practitioner when he has to select the criteria typifying non-random populations samples. His choice is made in the light of the aims of the survey concerned and has to take account of two imperatives: 1. the criteria must be readily, identifiable by the field interviewers; 2. they must bear causal relationships to the variables at issue. The second part tries to demonstrate how a more precise formulation of these requirements enables the problem to be seen in terms of statistical analysis, then have DSP libraries can help to find a solution. An account is then given of the programmes which were used for fixing the quotas involved in a 5,000-household survey of electrical goods ownership: a data consolidation programme followed by a segmentation programme. Brief notes on the results of this first experiment hope to stimulate interest in the pre-survey application of methods which have hitherto been deployed mainly at the post-survey stage.
- This could also be of interest