The Demopædia Encyclopedia on Population is under heavy modernization and maintenance. Outputs could look bizarre, sorry for the temporary inconvenience
Multilingual Demographic Dictionary, second unified edition, English volume
16
Disclaimer : The sponsors of Demopaedia do not necessarily agree with all the definitions contained in this version of the Dictionary. The harmonization of all the second editions of the Multilingual Demographic Dictionary is an ongoing process. Please consult the discussion area of this page for further comments. |
Go to: Introduction to Demopædia | Instructions on use | Downloads |
160
Sampling procedures ^{1} are used to obtain information about a population from part of the population only, instead of having to study every person (110-2). The part of the population studied is called a sample ^{2}. A population is a collection of elements ^{3} which are the object of the investigation. A sampling unit ^{4} may be an element or a group of elements of the population and is used for selecting samples. In demographic samples the elements are usually individuals (110-2), families (115-1), or households (110-3) and sampling units may be individuals, households, blocks of houses, municipalities or areas. The sample will consist of a number of sampling units selected in accordance with a sampling scheme ^{5} or sampling plan ^{5}.
161
A sample whose elements are selected by a chance process is referred to as a random sample ^{1} or probability sample ^{1}. If a complete list of sampling units is available, this is called a sampling frame ^{3}. In simple random sampling ^{4} a proportion of sampling units is selected from the frame at random ^{2}. This proportion is called the sampling fraction ^{5} or sampling ratio ^{5}. Systematic samples ^{6} are drawn systematically ^{7} from a frame in which the sampling units are consecutively numbered. The sample is selected by taking the n^{th}, (n + s)^{th}, (n + 2s)^{th}, ..., etc. unit, where n is not larger than s and is selected at random. In cluster sampling ^{8} population elements are not drawn individually, but in groups which are called clusters ^{9} .
- 2. Random, adj. - randomness, n. - randomize, v,
162
In stratified random sampling ^{1} the population is divided into a number of strata ^{2} which are in some sense more homogeneous (134-4) than the population as a whole with respect to the characteristics studied, and a simple random sample (161-4) is drawn in each stratum. Variable sampling fractions (161-5) may be used in the different strata. Multi-stage sampling ^{3} is a method where the selection of the sample is carried out in several stages. A sample of primary units ^{4} is first selected and each of these units is then regarded as a population (101-3) from which a sub-sample ^{5} of secondary units ^{6} is selected, and the process may be repeated. When there is no good sampling frame, a sample of areas delimited on a map may be selected: this procedure is called area sampling ^{7}.
- 1. Stratify, v,: divide into strata (plural of stratum) - stratification, n.
163
In probability sampling (161-1), chance methods are used to obtain a representative sample ^{1} i.e., a sample which is a faithful reflection of the population with respect to all the characteristics under investigation except for random fluctuation. In quota sampling ^{2}, on the other hand, the sample is purposely selected so as to reflect the population in certain characteristics, and each interviewer (204-2) is given a quota ^{3} of different types of sampling units which are to be included in his/her sample. Within the limits of the quota the interviewer is free to select the sampling units.
164
A population parameter ^{1} is a numerical value that characterizes a population. Statistical estimation ^{2} is the name given to the procedure by which the values of such parameters are estimated from the sample. Such estimates are subject to sampling errors ^{3} and a measure of the magnitude of the sampling error is generally given by the standard error ^{4}. Sometimes a confidence interval ^{5} is associated with an estimate to show the limits within which the estimated quantity may be expected to lie with a pre-determined probability. A difference between two values is referred to as a significant difference ^{6} when the probability that it is due to chance is less than a given value which is called the level of significance ^{7}. Thus a difference would be significant at the 5 percent level if the probability that it could have arisen by chance is less than 0.05. In addition to sampling errors, observation errors ^{8} or response errors ^{8} also affect estimates. These errors usually include interviewer biases ^{9} which are systematic errors introduced by the interviewers when the basic data are collected.
* * *
Go to: Introduction to Demopædia | Instructions on use | Downloads |