- 1 Objectives
- 2 Introduction
- 3 Full population studies: the census
- 4 Studying part of the population: sampling
- 5 Random sampling
- 6 Non-random samples
- 7 Sources of error in sampling
- 8 Heading
- 9 Heading
- 10 Heading
- 11 Conclusion
- 12 References
- 13 Discussion questions
- 14 Problems
- 15 Glossary
Ideally, when doing social scientific research we'd study the full population of interest. However, given the limited resources available to political scientists for research (for example, the annual budget for the political science research provided by the U.S. National Science Foundation is around $10 million<ref>Drezner, Daniel W. (October 7, 2009). "Tom Coburn picks on political science.". http://drezner.foreignpolicy.com/posts/2009/10/07/tom_coburn_picks_on_political_science. Retrieved April 21, 2012. </ref>), realistically large-scale research must make do with a subset of the population or a sample. Careful sample design, however, is necessary to ensure that our sample will be representative of the population of interest.
Full population studies: the census
With relatively small populations, or for specialized purposes, it may be reasonable to study the full population. Such a study is referred to as a census. The most famous examples of censuses are the regular censuses conducted by most countries to catalog the characteristics of their populations, such as the United States Census conducted every ten years and the Canadian census conducted every five years. In democratic countries, these censuses are important because they determine the basis of representation in national and subnational legislative bodies; they can also be important for the allocation of societal resources and for understanding the nature of social and economic problems faced by the population at large.
However, social scientific researchers almost never have the resources sufficient to conduct such a census; for example, the 2010 U.S. Census cost an estimated $13 billion U.S., or $42 U.S. per respondent, and required the hiring of thousands of enumerators and data processing employees. A census may be practical in more limited settings, such as a small community or an institutional setting (such as a university or college) with a relatively small population. A census may also be an appropriate approach to comparing the characteristics of large political units (for example, studying countries or subnational political units such as states and provinces) or the members of political bodies (such as a parliament or other legislature).
Studying part of the population: sampling
For relatively large populations where a census is impractical, studying only a part of the population is likely to be more feasible. However, the researcher must take care to ensure that the sample is likely to be representative of the population being studied if s/he is interested in inference, or learning about the population as a whole.
Social science researchers generally divide samples into two broad categories: random and non-random samples.
Simple random samples
Sources of error in sampling
- [[Def: ]]
- [[Def: ]]