Research Concepts

Sample Seledtion

Since a sample involves incomplete information, there is always a danger of introducting bias in sample selection. The primary strategy to avoid bias is randomization: every member of the population has an equally likely chance of being selected for the sample.

In principle this sounds easy: it is like conducting a huge lottery in which the "winners" get to be in the sample. In practice this is harder than it looks. The draft lottery of 1970 is an example of a poorly executed attempt at randomization in which dates earlier in the year were far more likely to have low draft numbers than those later in the year.

One approach to randomization is to construct a list of every member of the population, assign a random number to each member, then use a scheme to select the sample such as the lowest 500 random numbers. Of course to do this we must have a list of everyone in the population.

Systematic selections, such as every 1000th member of the population, might result in bias since there might be unexpected periodicities in the original list that favor one selection over another. That is why randomized lists are used.

Randomization recognizes that errors will necessarily occur in sampling. By randomly selecting the sample, the chance of introducing systematic error or bias is reduced.