ESP 178 Applied Research Methods
1/22 Class Exercise: Sampling
Introduction
Sampling is a critical element of the research process. It is generally not feasible to study the entire population
we are interested in, so instead we study a sample drawn from that population. As long as the sample is
representative of the larger population, we can assume that the findings for the sample hold for the larger
population as well. The quality of the sample determines the “generalizability” or “external validity” of the
results. There are now well-established ways of ensuring a high quality sample.
Questions
1. Let’s think about our study of biking in Davis. Let’s say we’re planning to do a survey that will let you
quantify levels of biking and characteristics that might influence biking. For the study, individuals are the unit
of observation and adult Davis residents are the target population. We have lots of different options for
sampling.
a. A good starting point is to think about possible sampling frames. Remember that the sampling frame is the
list of elements for your population of interest, in this case, individuals living in Davis. Here are a couple of
classic sampling frames you might use:
- Phone listings: What systematic biases might this sampling frame introduce into the study? What might
you do to reduce this bias?
- Marketing listings (available for a small fee from commercial providers): What systematic biases might
this sampling frame introduce into the study? What might you do to reduce this bias?
- Addresses (available through the Census bureau and local government sources): What is the unit of
analysis for this sampling frame? If we still want to study individuals, how might we choose them,
using this sampling frame? How can we ensure that we end up with a random sample of individuals?
b. Let’s say you use the address list as your sampling frame. What approach might you use for simple random
sampling? What approach might you use for systematic random sampling?
c. Researchers use stratified sampling for three general purposes: when there isn’t one sampling frame for the
entire target population but there are separate sampling frames for different subsets of the population (usually
proportionate), to create more homogenous subsets of the target population and thus tighten the confidence
intervals (usually proportionate), or to ensure sufficient numbers of respondents for groups that are a small
share of the population (usually disproportionate).
What characteristic might you want to stratify the population by? Why?
What sampling frame would you use for each strata? Think carefully: is there a list of the elements of
strata that you can get access to?
For proportionate sampling, how do you determine what share of the sample falls into each strata?
For disproportionate sampling, how do you determine what share of the sample falls into each strata?
Which one – proportionate or disproportionate - would you want choose for this project? Why?
