This lesson is an introduction to basic sampling designs. It does NOT cover these topics in great detail, but instead provides on overview of this broad statistical field.

After completing this lesson you should be able to:

The design used when collecting data is not only the first step in any experiment, but it is a crucial step in understanding the inference of a given study. There are five fundamental questions you should consider when designing an experiment or reviewing a published experiment.

1. What are the objectives of the study?
2. What are the variables of interest?
3. What is an expectable sampling design?
4. How should the data be collected?
5. How will I analyze the data?

Understanding the problem which the study is attempting to address is the first step you should use when designing your own experiment or reviewing a published experiment.

Example

A land management agency wants to assess the publics perception of wildfire use policies in order to increase the amount of wildfire use they have. Therefore the agency must determine which aspect of wildfire use determines a person’s perception of wildfire use.

Surveys

Surveys are a passive form of data collection, where the goal is to gather data on existing conditions, attitudes or behavior. Therefore in our example the agency may want to use a survey to sample current residents who live in areas which are currently using wildfire use in their forests.

Experiments

Experiments or scientific studies are an active approach to data collection. In conducting an experiment one varies some environmental condition and then studies the effect of that condition on the outcome. In our example, the agency could simply increase the amount of wildfire use they have in some areas and assess whether the opinion of the people in those areas has changed.

It should be noted that when conducting an experiment as many factors should be held constant as possible.

Example

We could measure the effects of a fertilizer on tree growth on a national forest, but we would have little control over temperature, humidity, insects etc… where as if we conducted the same experiment in a green house we could control all of these variables.

Surveys are used all the time in popular media. For example, a public opinion poll. They are also used by scientists, engineers, and corporations all the time. One of the most common surveys is the census conducted every ten years.

Simple Random Sample

One method of sampling is a simple random sample. Using this design we select n units from the population in such a way that each sample has the same chance of being selected. In other words, every experimental unit has the same probability of being selected to be surveyed.

Example

If we wanted to collect information about public opinion of the Forest Service in a particular town we could simply get a list of all the residents of that town, and randomly choose names from the list to take our survey.

Stratified Random Sample

Another approach, which is similar to simple random sampling is stratified random sampling. Using this approach we divide the population into two or more groups based on some factor, such as age or income. Once the population is stratified we then use the principals of a simple random sampling design in each stratification.

In the example above, we may be concerned about the opinion of men and women about the forest service in this town. So we would randomly choose names from a list of all the men and all the women who live in the town to take our survey.

Systematic Sampling

Systematic sampling is another technique often used when conducting a survey. Under this design we may sample every tenth person from the population. Often time’s systematic sampling will be called random systematic sampling. This implies that a random starting point was used and then every tenth person was sampled.

Example

Measurement Problems

In the case of measurement problems, we are mostly talking about issues with the wording of surveys, and the way a question is phrased. Specific problems can arise when a interviewee can not remember the answer to a question, or when the question is designed to lead a person to answer in a way that may not represent their opinion, and often times a question is unclear. There are many texts and courses which are designed to provide more information on how to create and conduct surveys.

Nonresponsive Participants

Designing a good sampling strategy can be a very challenging task for researchers, and agency personnel who are conducting a field inventory. Just as with surveys, you must first identify what you will sample and what your population is.

Example

1. First, the experimental units are easy to locate since they are evenly spaced.
2. Second, they appear to be more representative since they are evenly spaced across the population.

However, the disadvantages are that it is hard to estimate variance or standard error for one systematic sample and that it is possible the accuracy could be very low if the sampling coincides with a periodic variation inherent in the landscape.

Although this sampling design has been widely used it is best suited for times when only an estimation of the mean is needed. If an estimation of precision needs to calculated it would be best to use a random sampling design

Example

We could divide a watershed up into vegetation classes or into treatment blocks. Once the strata are identified at least two samples will be collected from each strata or subpopulation.

The advantage of this sampling design is that you can have a more precise estimation of the population mean compared to a simple random design of the same size. It may also be beneficial to look at each subpopulation. For example if each stratum is a different vegetation type we may have different long term goals and management tools to consider when treating the population.