When we think of sampling, many of us think of collecting pastries and heaps of macaroni and cheese from a buffet line. In research, selecting samples is not quite as appetizing. It requires more skill and precision than simply selecting what you want, and there’s no one formula for drawing the best samples. Instead, researchers employ several different strategies depending on their project parameters: time, budget, and study population.
When you want a sample that looks just like your study population (or a plate that looks a lot like a buffet), you’ll want to use one of two types of probability sampling: random sampling and stratified sampling.
Random Sampling: This is what happens when you let your nephew go to the buffet unattended. Each item has an equal chance of being selected and mashed together with a fork. Achieving randomness in research is much harder than you might think. For each person of interest in a population to have an equal chance of being selected, you cannot simply stand outside of Starbucks and pick the first five people who agree to take your survey—it would skew toward people who drink coffee at a certain time of day. Instead, you must employ methods such as random digit dialing or calling every 10th person in the phone book, to ensure randomness in the selection process.
Stratified Sampling: To collect a stratified sample, you first divide the population into categories-much like you might divide your plate at the buffet, making room for a meat, a bread and vegetables (I know I always spare half my plate for salad to cut calories, only later to find that I’ve drenched it in ranch). With a study population, you might create strata based on characteristics such as gender or social class before randomly sampling.
What are the benefits of probability sampling?
- Avoids sampling bias
- Allows researchers to estimate the amount of sampling error
What are the drawbacks of probability sampling?
- Time consuming
Although probability sampling has the advantage of creating a well-balanced data set (or diet), non-probability sampling is more frequently used for a variety of reasons. Quota sampling, purposive sampling and convenience sampling are the three flavors of non-probability sampling.
Quota Sampling: In a quota sample, the proportions of a stratum are set at certain levels. Like, I’m going to take five heaping tablespoons of mashed potatoes because I want to study them more intently, even though this means that an item that constitutes 1 percent of the food offerings takes up half the plate.
Purposive Sampling: A purposive sample is drawn for a specific need or purpose. Let’s say Aunt Margaret invited me out to dinner, but I’ve already eaten. I only select dessert items. These items are not wholly representative of the restaurant’s offerings, but serve my sweet tooth (my life’s guiding purpose).
Snowball Sampling: Sometimes you can’t find what you want despite being at a buffet—maybe hot sauce. You ask the Chef to bring out the Tabasco sauce and then whisper to the bottle of Tabasco to bring his friends A1 and Sriracha to the rave on your taste buds. In social science research, certain types of participants are equally difficult to locate—vegetarian hunters, Nyquil addicts, adult male fans of Twilight. When you find one, your best chance of finding another is to ask the former if he has any friends.
Convenience Sampling: As the saying goes “take it if it’s easy and if it’s easy, take it.” Sometimes in life, you need to take what you can get. To finish up with the buffet metaphor, if there’s an old man ahead of you going too slowly, you may just take the first five items and go scarf ‘em down. Researchers are frequently under similar time constraints, which explains why college sophomores in large lecture classes are the most studied people on Earth.
What are the advantages of nonprobability sampling?
- The sample can have the same attributes as the population being studied
- Faster, easier, and cheaper than probability sampling
What are the disadvantages of quota sampling?
- Don’t allow researchers to make generalizations or probability statements about their statistics
Food for thought: Now that you know the different styles of sampling in social science, which do you use when picking friends?