Producing Data

3. Sources of data

3.2 Generating your own data

3.2.2. Generating data from observations

iv. Types of sampling used to generate data

There are two basic types of sampling used to generate data - probability sampling and non-probability sampling.

1. Probability sampling

In probability sampling each person or unit in the population has a known or calculable chance of being included in a sample.
The four main types of probability sampling are:

1. simple random sampling (SRS),
2. systematic sampling,
3. stratified random sampling, and
4. cluster sampling.

a. Simple random sampling (SRS)

Simple random sampling means that each and every unit has the same probability of being chosen into a sample. This suggests the common activity of \drawing numbers or names from a hat. More commonly each member of the population is identified by a number and random number tables or programs are used to select the required sample size.

From a statistical perspective the benefits of simple random sampling are:

• It eliminates volunteers.
• Each unit/member of the population has the same chance of being selected.
• Because the sample is selected by chance, the sample should contain members with characteristics similar to the population as a whole.

One of the problems of simple random sampling, however, is that you need a complete list of the population, called a sampling frame, so that you can select, at random, the required number of units for the sample. Sometimes such a list is not available. Also, if sampling shoppers at a shopping centre, for instance, it would be awkward and impractical to attempt to collect names and telephone numbers to choose the sample from.

b. Systematic sampling

Practically speaking, systematic sampling this is easier to implement - a starting point is chosen and then a sample member is selected at fixed intervals. In selecting a sample of size 100 from a population of size 6000, we would select a random number between 1 and 60 for a start point, and then choose every 60th item until we achieve the desired sample size of 100. The choice of 60 is made because the population is 60 times larger than the sample size.

c. Stratified random sampling

Stratified random sampling reflects the fact that populations are often divisible into layers or strata. In an attempt to create representativeness, the sample is chosen similarly to the population - the same number of strata, and a proportionate number of units in each stratum. If conducting a survey on attitudes to improving women’s support services on a campus, it would be wise to construct a sample which reflected the male / female strata proportion in the campus population. In this method every member does not have an equal chance of selection, but the probability of selection is calculable.

SCENARIO [2] At the University of Wollongong about 20% of the engineering students are women. The Faculty of Engineering plans to poll a sample of 200 engineering students about the quality of student life.

d. Cluster sampling

With cluster sampling, a population is divided into groups (particularly geographic), then some of the groups are randomly selected and then either an SRS or a census is conducted in these chosen groups. Again, every member does not have an equal chance of selection, but the probability of selection is calculable.

2. Non-probability sampling

With non-probability sampling, not every unit has a chance of selection in the sample and the process involves some amount of subjectivity instead of following predetermined, probabilistic pathways. This can be useful in small scale exploratory studies where we wish to gain great familiarity with the population rather than to reach statistical solutions.

The three main types of non-probability sampling are:

1. convenience sampling,
2. purposive sampling, and
3. judgement sampling.

a. Convenience sampling

Convenience sampling means that members of such samples are chosen mainly because they are readily available and willing to be involved - hence there is a saving of time and money.

Such samples might not be representative of the population and so it might be difficult to make conclusions about a population based on this type of sample. If your sample is made up of volunteers, then it is likely to be biased because the volunteers may be actively supporting/promoting a point of view. Television stations often use convenience sampling when they ask viewers to phone in responses to a question. Also the 'person in the street' interview is often conducted in a street near the TV studio, or even in the foyer of the TV station amongst waiting audiences.

Another limitation of this method is that it does not set out to completely identify the population being studied.

b. Purposive sampling

With purposive sampling members are chosen deliberately because they are not typical of a population. The maker of a new scientific calculator could give samples of the new item to students in an engineering faculty - if they cannot easily understand it, the general population could not either. The engineering students are a purposive sample.

c. Judgment sample.

Judgement sampling is used when the researcher believes the members to be representative of the population. The representativeness of such a sample is only as good as the researcher’s ability. Many business decisions are made by managers who operate this way without quantitative support.

SCENARIO [2] At the University of Wollongong about 20% of the engineering students are women. The Faculty of Engineering plans to poll a sample of 200 engineering students about the quality of student life.

There are other forms of sampling. Many of them can be thought of as a combination of purposive and random sampling [3]. You want your samples to be representative of the population. If there are any limitations with your sampling procedure, it is important that you acknowledge them in your report because they can influence the validity and reliability of your results.

Next - Menu