Abstract:

Sampling theory is well developed and well known. The most satisfactory methods are also simple. Unfortunately the theoretically sound and simple methods can seldom be applied in practice in sampling human population. It is often impossible to find sampling frames of exactly the kind required, and the cost of using the best procedures is frequently prohibitive, even when frames can be found. To overcome these difficulties some reasonable short cuts have been developed. Stratification and stage sampling are among the more satisfactory techniques. Considerable ingenuity has been used to find frames for unusual populations and the use of correlates has been of value. For some purposes the known sampling distributions of certain statistics (like Pearson's product movement and chi-squared) can be compared with distributions observed in practice. For example, these comparisons have been used to detect cheating interviewers. Fundamental probability sets can provide theoretical frames from which theoretical samples can be drawn and the products represented by the sample produced to the specification. This article describes some unusual sampling problems and the ways in which they were dealt with.