Selecting Samples (Continued)
Now that you have some idea of how probability theory works. There are different ways of actually selecting samples.
I. Types of representative samples.
A. Random samples
Truly selecting items at random.
Examples: Random digit dialing.
Advantages: Representative sample is obtained
Disadvantages: Sampling error is higher than
other methods.
B. Systematic Sampling
Every Nth item is selected.
Very similar to random sampling, if the list itself is random to start with.
Advantages: Very easy to do. Don't have to generate random numbers.
Disadvantages: Periodicity may be a problem.
If items on the list are arranged in some
systematic order and somehow you fall into this pattern when
selecting your sample.
C. Stratified Sampling
Mixture of random and systematic sampling.
That is, you intentionally try to identify various groups or sub-parts within the population and then you sample elements randomly from within each sub-part of group.
The purpose is to create sub-groups within the larger population and then sample from those sub-groups.
For example: Many polling firms, stratify or group people according to their sex and then they sample people from each group.
Advantages: Produces more representative sample. You guarantee that your sample will be representative along some criteria (the one you used to classify or group people).
Reduces the sampling error involved. As long as the criteria used to create sub-groups is related to the other variables that are ultimately going to be studied.
Disadvantages: Prior knowledge about population is needed. You need to have extensive knowledge about your population in order to divide them into groups.
Also very expensive.
D. Multistage Clustering
Method where multiple samples are taken in successive iterations.
Let's say you want to randomly sample all TV shows for a year.
In Santa Barbara they are randomly selecting TV shows for an entire year.
How would you go about doing that?
If you wanted to use simple random sampling, you'd have to list each show for the entire year, and then select from those. (This isn't very easy to do, if possible to do at all)
Can you think of another way to doing it?
What they did, is they randomly select days, not TV shows for the year. It is very easy to list all of the days of the year.
Next, once you select days, they then randomly select hours from each of the day taken. (again it is very easy to list all of the hours in a day).
Next, you list all of the shows for the hours that were selected. And they randomly select those shows for those hours.
Essentially, you use random selection repeatedly, until you finally select the items you really want to study.
Advantages: Very easy to do, great for complex and large populations (where it isn't feasible to list every item).
Disadvantages: Sampling error is increased.
Every time you sample, you produce some sampling error. So, each
time you sample, you increase your error a little bit. In other
words, your sampling errors accumulate every time a sample is
taken.
Overall advantages:
These methods are very useful, because they allow you to make claims about large population, based upon the observations of only a select few.
Examples when should be used?
Trying to gauge public opinion. (instead of just person on the street)
Whenever you want to make very broad and
general claims: "There is a lot of violence on TV"
II Types of Non-representative Samples
Many times when selecting people or things to study, we do not employ methods that produce representative samples.
In fact, most research within the field of communication does not employ representative samples.
A. Purposive Sampling
Sometimes researcher simply wants to study a particular group, and is not interested in generalizing to larger populations.
B. Convenience Sampling.
Items are selected because they are there.
Examples:
People on the street interviews.
Those people who stop you in the mall.
Most social science research.
When convenience sampling is employed, who do the results represent? They only represent the actual people that you observed. They do not necessarily generalize beyond the sample.
However, that is why we replicate studies. Because if the same results are obtained, at different locations, using different people, then we can place more faith in those results.
So, a single study, using convenient subjects,
should not be generalized to the population at large.
C. Quota Sampling
Very similar to stratified sampling in that the population is classified into various sub-groups.
Moreover, like stratified sampling a proportional amount of individuals in each sub-group is selected to be in the sample.
However, the key difference between quota and stratified sampling is in how people or items are selected. In stratified sampling the elements are chosen at random (where everyone has an equal chance). In Quota sampling, the individuals are selected in whatever way the investigator chooses.
It is in this selection process that biases are
introduced into the sample.
Overall advantages:
These methods are very easy, quick, and cheap in comparison random sampling techniques.
Overall disadvantages:
When you do not employ random sampling techniques, you do not learn about the overall population. You can't make claims about the larger population. You can only make claims about the people you actually observed (your sample).
Examples when should be used?
When testing a survey before using it with a representative sample.
When you are not interested in generalization. When you want to study a very specific or particular group.
When you are doing exploratory research. Just trying to figure out what is going on at a basic level.
When you know that others will replicate your work.
When you don't have the resources to use random sampling techniques.