Importance of Sampling in Statistical Analysis

The validity of a statistical analysis depends on the quality of the sampling used. The two most important elements are random drawing of the sample, and the size of the sample. A small sample, even if unbiased, can fail to include a representative mix of the larger group under analysis. A biased sample, regardless of size, can lead to incorrect conclusions.

When a survey intends to depict trends in some particular group, its sample must come from that group. Screening questions can sort out qualified and unqualified respondents for some known criterion, such as whether they are consumers of a certain product or service, or whether they are parents or household heads. But sometimes the screening must deal with probabilities, such as whether the respondents are likely to vote or to buy something. In these cases, the sample should be chosen based on criteria like past voting participation, or previous purchase of a similar product or service.

Convenience sampling -- such as door-to-door or "man on the street" interviews -- is drawing respondents who are convenient to the survey taker, perhaps people at an event or particular place. While quick and easy, and seemingly random, it isn't a truly random sample because only those who are eager to respond will be included. People with strong opinions will be over-represented in the sample. The nature of the location can introduce bias, such as a political event where a disproportionate number of respondents express some particular opinion.

Systematic sampling draws one survey subject out of a given number from some group, such as one out of every 50 names in a telephone book, or people in a particular place, or people in some defined group. It isn't as subjective as convenience sampling because it doesn't accept just anyone who wants to express an opinion. But it isn't completely random because only the people in the book, chosen place or group can be in the pool. If the survey scope is limited to those groups, then its results can resemble those of a random sampling method.

The point of random sampling is to avoid bias from factors such as the eagerness of respondents to express an opinion, a limited base for the sample, and chance availability of respondents. It can require persistence in attempting to contact selected individuals; otherwise, as with convenience sampling, only those easily available or eager to express an opinion will be counted. A truly randomized sample of a population under study offers a glimpse of the population as a whole.