A sample of a population is drawn when the population is either too large or unavailable for study in its entirety. A sampling strategy is the process of identifying your population and then determining how to best select a sample from it. Choose a feasible plan and recognize limits. Populations do not have to be people. They can be objects such as businesses, countries, parks, etc.

**Representative/random**

The list below several types of fairly chosen samples of significant size that can be considered representative of the larger population to which they refer. Other common sampling types are random, probability and quantitative. To qualify as a random sample, each unit of the target population must be given an equal chance of being selected, much the same as in a lottery draw. Selection is based on the laws of probability. Randomly drawing the requisite sampling proportion from the target population using a sampling frame makes inferential statistical analysis and tests of statistical significance possible. Using these, you can calculate the probability that the sample statistic derived from a relatively small sample would apply to the entire population if a census were taken. Estimates of possible errors (random sampling errors) can also be estimated using probability or chance.

**SRS**

This stands for Simple Random Sample, the best known of the four basic types of random sampling techniques. SRS gives each unit in the sampling frame (the list from which the sample is drawn) an equal chance (probability) of being selected. Drawing names from a hat or using the table of random numbers are two ways to select a simple random sample. It is only feasible if a list can be generated. For confidentiality purposes and other reasons, lists are often not available.

**Systematic**

Systematic random sampling is most useful when the sampling frame cannot be easily listed in a document. As a probability-based selection technique, it guarantees that each unit in the sampling frame is given an equal chance of being selected. The technique involves selection of units at regular intervals with a random start. A perfect example is systematic sampling library books in the library’s collections. A hundred books could be selected in this way by starting with a random number and selecting every twentieth book until you get a hundred books. Human selection bias is avoided and random sampling error can be calculated based on probability or the random chances of selection bias.

**Stratified**

This technique is combined with one other random technique, either simple random or systematic. Proportions, both directly proportional to the target population (30/70) or equally proportional groupings (50/50), are set in advance and then within each of those proportional groupings, either simple random or systematic sampling is applied.

**Multi-stage cluster**

Usually called multi-stage cluster random sampling, this is a combination of the three other random sampling techniques. It is done in multiple stages and is most useful for drawing a random sample from a very large, diverse sampling frame. Taking a random sample of adult Canadians requires multiple steps to select a sample that represents all the regions and various sectors of Canadian society. The map would most likely select an equal proportion of adults (stratified) from each of the ten provinces and three territories, and then by rural, suburban and urban areas (stratified), and then broken down into neighbourhoods (systematic or simple random) and then finally households (systematic or simple random until the requisite sampling proportion is drawn). Most country-wide polls select using this multi-stage cluster sampling technique and the sample size is usually around 1200-1500.

**Non-representative/non-random**

The list below describes samples that are small and selectively chosen and do not represent the population from which they are drawn. Other techniques from this general category are non-random, non-probability or qualitative. Non-random selection techniques, which selectively draw the sample from the target population, are very popular and practical. Non-random sampling is typically used in most research studies, as randomization is difficult, time-consuming, expensive and sometimes not feasible given the parameters of most student research timetables and budgets. It is often impossible to locate a statistically significant portion of very specialized populations, such as female mountain climbers or musical prodigies.

**Convenience**

The name itself provides an accurate picture of this type of non-representative sampling as “easy” or “conveniently available” for the researcher. Sources of selective bias must be recognized by the researcher. If the researcher selects a convenience sample consisting of a few close friends, then the researcher is obliged to reveal potential sources of bias such as similar educational and socio-economic backgrounds. Accidental samples are closely related to convenience and often used by historians and anthropologists. Often only a few artifacts survive the tests of time, so whatever artifacts remain must suffice, thus making results highly tentative and specific to the reasons for the survival of these few remaining traces.

**Purposive**

Purposive sampling a common non-probability sampling technique useful for obtaining access to individuals (not always people) that are not easy to locate in the general population. Also referred to as judgmental sampling, it gives the researcher the chance to pick and choose units that qualify when there is a limited or specified pool to draw from. If your target population were amateur chess players, for example, you would be well-advised to visit locations such as a college or university chess club, where you may “purposively” go to the club area to locate a few members of this community.

**Snowball**

Snowball sampling is a practical qualitative sampling technique for difficult-to-find research subjects or units. It involves accessing subjects or units through network connections. One person recommends other people for inclusion in the sample and it grows exponentially from there. Studying hobby farmers is made easy; once you locate one hobby farmer, you can ask them to recommend a few other hobby farmers they know. Gaining access through ready-made networks could lead to systematic bias, so be aware that no generalizations to the population of hobby farmers can be made, only surmised or suggested.

**Theoretical**

Theoretical sampling is a typical qualitative sampling technique that fits well with an open-ended, exploratory design. The actual number and type of the sample is determined in an ad hoc way over the course of the research study. The researcher has a clear goal but does not start off with a pre-set, precise sampling plan. Instead, it sets a tentative plan to sample a certain number of units until no new information is being revealed and a saturation point has been met.

**Quota**

This is usually combined with any of the other non-representative sampling techniques to ensure representation by gender, political party affiliation or any other criterion that may be of relevance to your study.