What are the disadvantages to the classic experiment?

The primary disadvantage of an experiment is that the controls imposed by a researcher to control for rival causation may create artificial conditions that alter the ability for the results of the study to be generalized to other populations. This is known as artificiality. Other disadvantages include difficulty in mimicking real-life situations in experimental situations, difficulty in retaining subjects, increased ethical considerations, researcher bias, higher risks to internal and external validity, and more difficult generalization of findings.

Differentiate between probability sampling and non-probability sampling.

Probability sampling refers to samples that permit estimation of the likelihood of each element of the population being selected in the sample. These include simple random samples, where each element has an equal likelihood of being placed in each group, stratified random samples, which include placement in a group based on a single demographic characteristic with all other variables as random and equal, cluster sampling, or grouping based on clusters, rather than random assignment, systematic sampling, or the sampling of a specific single occurrence within a group, such as using every eighth soil sample, and multi-stage sampling, or a combination of the above methods. Non-probability samples refer to any sample that violates the concept of equal likelihood of placement. These include quota samples, or stratified samples for a single purpose, accidental samples, such as 'person on the street' sampling, and purposive samples, or samples chosen for a specific variable.

What is the purpose of probability sampling?

Probability sampling provides ways to select samples in which every unit in the test population has an equal chance of appearing in a single sample. This is necessary, since random sampling, and equality of sampling is needed in order to generalize findings to the wider population. By using probability samples, researchers are ensuring a random selection of the population, helping to raise confidence limits, so degree of error is smaller.

Differentiate between the types of probability sampling...

Stratified random samples, which include placement in a group based on a single demographic characteristic with all other variables as random and equal, is useful in mimicking particular populations, such as the number of females in a given area. Cluster sampling, or grouping based on clusters, rather than random assignment, is useful in studies where the population is widely dispersed. Systematic sampling, or the sampling of a specific single occurrence within a group, such as using every eighth soil sample, is useful in that each sample is chosen without regard to any other characteristic. Multi-stage sampling, or a combination of the above methods, is useful in that by combining a variety of sampling methods, one can ensure randomization while still controlling for variables.
What are the disadvantages of non-probability sampling techniques?

First, non-probability sampling techniques do not allow for statistical probability calculations, which mean any results discovered cannot be generalized to the wider population. Often, these sampling techniques also rely on the researcher's view of characteristics, which may prove to be incorrect, making the sample useless. Further, the non-random sample is more costly, more time consuming, and less accurate for most forms of experiments, and is generally only used when specific populations are to be examined.

What is meant by the terms exclusive and exhaustive as they relate to variables?

When defining variables by breaking a group into subgroups, each grouping or coding category must have two properties. One is that each unit must only exist in a single group. This is known as exclusive. Secondly, each unit must exist in a group. This is known as exhaustive. So, in a sample of 50 units, each unit must be subjected to a single group and a single variable, and all 50 units must exist in a group. In some cases, this may require a group known as "other," if the identifiable characteristic is too vast to create a group for each unit possibility.