Psychology
Stratified Sampling
Stratified sampling is a research method that involves dividing the population into subgroups, or strata, based on certain characteristics such as age, gender, or socioeconomic status. Researchers then randomly select participants from each stratum to ensure that the sample is representative of the entire population. This approach allows for more accurate and reliable results by accounting for diversity within the population.
Written by Perlego with AI-assistance
Related key terms
1 of 5
12 Key excerpts on "Stratified Sampling"
eBook - PDFEducational Research
A Contextual Approach
- Ken Springer(Author)
- 2015(Publication Date)
- Wiley(Publisher)
Replication of findings (discussed in Chapter 1) helps to minimize the likelihood of sampling error distorting our understanding of some phenomenon. Stratified Sampling Stratified Sampling is a procedure in which strata (i.e., subgroups) of a population are identified, and random sampling is carried out within each subgroup. When distinctions between subgroups are important to a study, Stratified Sampling may be preferable to simple random sampling. For example, in a school district in which 91% of the elementary level teachers are female, a simple random sampling method, particularly for a small sample, might not yield any males. However, the representativeness of the sample would be improved if at least some males were chosen. In fact, the representativeness would be ideal if exactly 9% of the sample were male. This example illustrates one case in which a slight modification of pure random sampling would be desirable. In some cases, the goal of Stratified Sampling is for the proportion of each subgroup in the sample to be the same as the proportion in the population. This is called proportional stratified random sampling. In the previous example, researchers would randomly sample members of each gender until a balance of exactly 91% females and 9% males was obtained. The researchers might or might not choose to stratify their sample on the basis of additional variables. For instance, if each teacher is classified as either ‘‘experienced’’ or ‘‘inexperienced,’’ and it was found that 54% of female teachers and 61% of male teachers are experienced, the researchers would ensure that their sampling within each gender reflected these proportions. In other cases, researchers use nonproportional stratified random sampling when the goal is to simply represent different subgroups. This is a common approach when the researcher’s interest is to compare groups of individuals who differ in achievement, demographic background, or some other variable of interest.
eBook - PDF- R. D. Savage(Author)
- 2013(Publication Date)
- Pergamon(Publisher)
Then, to the subsample were added families selected on the basis of stratification so as to secure a better representation of the salaried and independent professional and business groups. The final sample of 30,000 was then chosen by the stratified method, the stratification being on the basis of the information obtained from the 250,000. From these 30,000 cases detailed information on expenditures will be secured. Few, if any, psychologists will ever be in a position to follow such an elab-orate sampling scheme. The study of Garrett, Bryan, and Perl [12] indicates one effective way of sidestepping the sampling problem. They used all the 9-, 12-, and 15-year-olds in a small city school system in a study which depended upon comparison of these three age groups in regard to the intercorrelations of tests. It cannot be argued that the found differences are due to selection. Another example of taking an entire school population as the sample is to be found in the study of intelligence and birth order by Steckel [43]. It should be noted, of course, that the nature of these studies is such that one would expect small variation in the results as one passes from city to city, and rather than attempt to draw individuals for represen-tative samples, these two studies depend more upon the choice of typical cities. A similar scheme was followed by Jones and Conrad [20] in their study of the growth and decline of intelligence. Their monograph not only illu-strates adequate methods for surmounting the difficulties of sampling, but also includes a thorough description of the group used and an enlightening exposition of possible selective factors in studies of this type. This study will repay careful reading by those who are interested in the sampling problem. 396 PSYCHOMETRIC AND STATISTICAL TECHNIQUES Perhaps no field of psychology is fraught with such complex sampling difficulties as those found in studies of race and nationality differences.
eBook - PDFSurveying Natural Populations
Quantitative Tools for Assessing Biodiversity
- Lee-Ann Hayek, Martin Buzas(Authors)
- 2010(Publication Date)
- Columbia University Press(Publisher)
108 FIELD SAMPLING SCHEMES Statisticians define this sampling scheme as Stratified Sampling . Occasionally, in the botanical literature, this type of sampling is called representative sampling . An advantage of this design is that the strata (or microhabitats, groups, or blocks) can be selected to cover important biofacies or habitats of interest (also called domains of study in the sta-tistical sampling literature), for which separate sample information is needed. The ad-vantages of Stratified Sampling are nearly always considerable, except when the popu-lation is known to be extremely homogeneous. The reduction of variance and the narrowing of the resultant confidence interval about the true mean density value are important attributes of this method. Once again, we stumble over the problem of the same words being used with quite different meanings in different disciplines. For a geologist or stratigrapher, the words stratum and stratification have an altogether different and distinct connotation. This can be particularly confusing because the stratum of the geologist is often a good unit upon which to base a statistically stratified sample. Nevertheless, we shall continue using the term Stratified Sampling as defined here because it has become an accepted part of the literature for non-geologic field sampling. Up to this point in this book, we have discussed methods that involve taking a sample from the entire population of interest. Stratified Sampling takes samples in-stead from subpopulations of the total target population. It requires prior knowledge of a limited (usually 2 to 5) number of zones, habitats, sections, subareas, subgroups, or biofacies (or even sexes, races, age classes, and groups defined by behavioral, morpho-metric, or descriptive variables) within the target population. Successful use of strati-fied sampling also depends on accurate identification of these strata, each of which is treated as a separate, independent population.
No longer available |Learn more- Frederick J Gravetter; Lori-Ann B. Forzano; Tim Rakow, Frederick Gravetter, Frederick Gravetter, Lori-Ann Forzano, Tim Rakow(Authors)
- 2021(Publication Date)
- Cengage Learning EMEA(Publisher)
Combining these two subgroup samples produces the desired stratified random sample. Copyright 2021 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it. C H A P T E R 5 \ Selecting Research Participants 119 Stratified random sampling is particularly useful when a researcher wants to describe each individual segment of the population or wants to compare segments. To do this, each subgroup in the sample must contain enough individuals to adequately represent its segment of the population. Consider the following example. A sociologist conducts an opinion survey in a major city. Part of the research plan calls for describing and comparing the opinions of adults from three different age groups: ‘Millennial’ (born 1981–1996), ‘Gen-X’ (born 1965–1980) and ‘Baby Boomer’ (born 1946–1964). If the researcher uses simple random sampling to select 150 individuals, the sample might contain only a few individuals from one (or more) of these groups. With only a handful of representatives of a particular group, the researcher could not make any definite statements about that group’s opin- ion, and could not make any meaningful comparisons with other groups. A stratified random sample avoids this problem by ensuring that each subgroup contains a predetermined number of individuals (set by the researcher). For a total sample of 150, the researcher selects 50 representa- tives of each of the three age groups. The main advantage of a stratified random sample is that it guarantees that each of the different subgroups will be well represented with a relatively large group of individuals in the sample.
- Timothy G. Gregoire, Harry T. Valentine(Authors)
- 2007(Publication Date)
- Chapman and Hall/CRC(Publisher)
CHAPTER 5 Stratified Sampling Designs 5.1 Introduction A Stratified Sampling design purposely partitions the target population, P , into two or more non-overlapping subpopulations, called strata , which are sampled separately. In this chapter we present the rationale for Stratified Sampling and estimators for the population total, y , and related population parameters, as well as corresponding estimators for the individual strata. A design issue which arises with Stratified Sampling is the allocation of the overall sampling effort to the various strata, a topic which is addressed following the section on estimation. The sections at the end of the chapter discuss matters related to Stratified Sampling: incorrect strata assignment, double sampling for stratification, and poststratification. 5.2 Rationale for Stratified Sampling Stratification is often motivated by a desire or requirement to estimate the total or average value of some attribute, y , for each stratum of interest. In the U.S.A., for example, natural resource and agricultural surveys are mandated and conducted by the federal government. Survey results are reported separately by state, so each state serves as a stratum, and sampling within each state is conducted independently of sampling in any other. Separate reporting is not necessarily contingent upon stratification because results may be calculated separately for each subpopulation of interest, whether or not the sampling is conducted separately. But stratification prior to execution of the sample is often advantageous since it enables the planner to customize the sampling design to the needs and features of each stratum. For example, among several states, one state may require greater precision in the estimation of its natural resources, so the intensity of sampling may need to be greater within that state.
eBook - PDF- Peter H. Rossi, James D Wright, Andy B. Anderson, Peter H. Rossi, James D Wright, Andy B. Anderson(Authors)
- 2013(Publication Date)
- Academic Press(Publisher)
In some cases, however, certain mathematical condi-tions will allow one or more of these steps to be carried out implicitly. There are three basic reasons why stratification is used in probability sam-pling. 1. The use of appropriate stratification may greatly increase sample effi-ciency (i.e., lower sampling variance). 2. By creating explicit strata, we may assure that certain key subgroups will have sufficient sample size for separate analysis. 3. The creation of strata permits the use of different sample designs for different portions of the population. Increased efficiency is probably the most common reason why stratifica-tion is used in the design of probability samples. When stratification is not used, the sampling variability of sample estimators is related to the variability that exists among the basic units in the population. For a given variable, this vari-ability is measured about the overall population mean. By dividing the popula-tion into strata, sampling error becomes a function of within-stratum variabil-ity. If the within-stratum variation is less than the overall variation, Stratified Sampling procedures will result in decreased sampling errors. Assurance of sufficient sample size for separate subgroup analysis is an-other important reason for using Stratified Sampling. A nonstratified equal prob-ability sample will yield sample cases from various subgroups in proportion to their distribution in the general population. Thus, with a nonstratified, equal probability design, we expect that if a particular subgroup represents 5% of the total population, it will constitute approximately 5% of the total sample. By creating separate strata consisting of particular subgroups of interest, we are free to increase or decrease the relative distribution of these subgroups in the sample. Stratified Sampling procedures also may be employed because of problems in obtaining adequate population frames.
- Frederic B. Mayo(Author)
- 2013(Publication Date)
- Wiley(Publisher)
Stratified random sampling brings the same method to a section of a population to make sure that all critical subsets or groups of a population are likely to be chosen as participants or respondents. For example, while a simple random sampling process might work for hotel guests, you might want to stratify the sample for reasons of traveling—such as leisure, busi- ness, or friends and family—or by gender or category of hotel guest—single traveler, couple, or family. In this way, you will get sample results from each of these groups that reflect the proportion of those sections of the larger population. In effect, you stratify the population into sections and sample according to those sections. Most researchers do it for efficiency and ease in applying various statistical methods. If you are interested in comparing the responses of regular restaurant cus- tomers to first-time visitors to a restaurant, then you will want to stratify the sample so that you randomize among those who are regular patrons and those who arrive for the first time or come casually but not regularly. Maintaining the randomizing process helps prevent adding bias to the results and ensures that each person—with their strati- fication group—has an equal chance of being selected as a respondent in the survey. Proportional stratified random sampling , like stratified random sampling, uses a random process for each stratification group but also ensures that the balance among the stratification groups matches the proportion of those stratifications in the popu- lation at large. It makes the statistical analysis easier, but it requires knowledge of the stratifications in the total population and the ability to ensure that the sample stratification groups match the total population.
eBook - ePubSampling Essentials
Practical Guidelines for Making Sampling Choices
- Johnnie Daniel(Author)
- 2011(Publication Date)
- SAGE Publications, Inc(Publisher)
- Stratified samples yield smaller random sampling errors than those obtained with a simple random sample of the same sample size, especially if optimum allocation is used. Stratification makes for a gain in precision, eliminating the variation of the variable that is used for stratifying. The amount of gain in precision is determined by the extent the within-stratum variances of the study variables are minimized and the between-stratum variances of the study variables are maximized. Stratification will yield a sample that is at least as precise as a simple random sample of the same size. If it is ineffective in increasing the level of precision, the results would not be worse than if simple random sampling were used.
- Stratified samples tend to be more representative of a population because they ensure that elements from each stratum in the population are represented in the sample. Sampling may be stratified to ensure that the sample is spread over geographic subareas and population subgroups.
- In using Stratified Sampling, advantage is taken of knowledge the researcher has about the population.
- If the stratification variable breaks up the population into homogeneous geographical areas, data collection costs may be lower than the data collection costs of sample random sampling.
- Utilizing Stratified Sampling permits the researcher to use different sampling procedures within the different strata.
- In using Stratified Sampling, a researcher may be created taking into account administrative convenience in carrying out the study. The researcher may take into account the clustering of the population in metropolitan areas, institutionalized segments of the population, and the distribution of data collection staff.
Compared to simple random sampling, weaknesses of Stratified Sampling include:- Stratified Sampling has a greater requirement for prior auxiliary information than is the case for simple random sampling. Information on stratification variables is required for each element in the population. Such information includes information on the proportion of the target population that belongs to each stratum; if optimum allocation is used, information on the variability of the variables of interest and information on the data collection costs are necessary for each stratum. Acquiring such information may be time-consuming and costly.
- Selection of stratification variables may be difficult if a study involves a large number of variables. These variables should be correlated with the variables of interests in the study.
eBook - ePub- C.A. Moser, G. Kalton(Authors)
- 2017(Publication Date)
- Routledge(Publisher)
Before going further, it may be useful to emphasize the respective roles of stratification and of spreading a sample over several stages. Although these two features of design fulfil quite different functions, the newcomer to the subject often has difficulty in distinguishing them. Continuing with our illustration, suppose that there are suitable lists of districts and of individuals within districts from which the sample of 2,000 adult civilians can be drawn. Let us look at four possibilities:Design I. Simple random sample
This would involve taking the list of adults for the entire country and extracting from it a sample of the required size, either by random numbers or by some systematic procedure.1Design II. Unstratified, two-stage sample
Sample I would be too thinly spread over the country to be practicable. We may decide instead to concentrate the sample in a certain number of administrative districts, selecting, say, 100 such districts from a complete list of districts. We then turn to the list of individuals for each of the selected districts, and from it select the required number (20 on average).Design III. Stratified simple random sample
Either because the regions are believed to differ markedly from each other with regard to the subject of the survey or purely for convenience, we may decide to stratify by region (say, with a uniform sampling fraction). The first step is to allocate to each region its appropriate number of interviews. If South-West England has 7 per cent of the adult civilian population of Great Britain, 140 out of the sample of 2,000 individuals would be selected from this region. Up to this point no sampling has taken place. Stratification is merely a way of dividing the population into a number of sub-populations and allocating the sample between these sub-populations. Once this is done the sampling begins, and with simple random sampling within strata we would select the number of individuals required in a region from the entire list for that region.Design IV. Stratified, two-stage sample
The stratification by region is as in Design III. But in order to concentrate the interviewing, we select a sample of administrative districts in each region and then a sample of individuals from each of the selected districts.
eBook - PDF- Ranjan K. Som(Author)
- 1995(Publication Date)
- CRC Press(Publisher)
310 Chapter 19 stratum in order to provide an efficient sample. Almost all nation-wide sample surveys, including those mentioned in section 1.1, use stratified multi-stage designs. For example, for the Na-tional Market Research Survey in Great Britain (1952), where the universe was the adult civilian population aged sixteen and over, the sample was stratified by factors such as geographical region, urban/rural areas, indus-trialization index and zoning, and was spread over four stages: adminis-trative districts, polling districts, households, and individuals (Moser and Kalton, Chapter 8). In the monthly Current Population Survey in the U.S., the sampling design is stratified three-stage; heterogeneous first-stage units are defined, comprising individual counties (or sometimes two or more ad-joining counties), and grouped into a number of strata; the first-stage units are sampled with varying probabilities; relatively small areal units of about two hundred households are the second-stage units; and the third-stage units are clusters of approximately six households, called segments: the sample contains 449 first-stage units, about 900 segments and about 50,000 households (Hansen et al. (1953), Vol. I, Chapter 12; Hansen and Tepping, 1969). In the Indian National Sample Survey (1964-5), in the rural sector there were 353 strata, comprising administrative units, and each covering an average of 1.2 million persons; for the socioeconomic inquiries, villages and households were the two stages of sampling, and for crop yield surveys, the four stages of sampling were respectively villages, clusters of plots, crop-plots, and circular cuts of radius 2' 3 or 4' (see Appendix V). In the Sample Survey on Goods 'Ifaflic Movement in Zambia (1967), the design was stratified two-stage, with proprietors of transport vehicles constituting the first-stage and the vehicles the second-stage units.
- Tarun Kumar Roy, Rajib Acharya, Arun Roy(Authors)
- 2016(Publication Date)
- Cambridge University Press(Publisher)
There are pockets in a population that may be difficult to access for various reasons. Separate strata, in a Stratified Sampling, can contain such segments that are relatively inaccessible and can, for example, be sampled with a lower rate (which can be taken care of with weighting at the time of the estimation). 2.3.7 POST-STRATIFICATION The principle of stratification can be employed at the time of estimation even when a sample is selected without the stratification. This is known as post-stratification. Let us elaborate this in the context of estimation of tobacco use in a population. Age is generally a variable that is correlated with the use of tobacco. Hence, it is desirable that a sample selection procedure should consider age as a stratification variable for a bias-free estimation of the use. The age distribution of the population is available but for some reasons stratification was not done at the time of sampling. The plausible reason for not considering age could be that the information was not available at the level of the sampling units. An EPSEM design is used without the stratification. Let, N i = number of persons from ith age group in the population N N i i = ∑ = total size of population n i = number of persons from ith age group selected in the sample n = total sample size P = proportion using tobacco in the population p i = proportion of persons from ith age group using tobacco in the sample y ij = a dichotomous variable indicating whether jth individual in ith age group is using tobacco ( ) y ij = 1 or not ( ) y ij = 0 p = overall proportion using tobacco in the sample Information on n i and p i can be easily obtained after interviewing the sample respondents. If the two distributions, N i and n i differ, instead of taking the simple average of p i , one can employ the principle of stratification and obtain the estimated p using equation (2.30) as p N p N i i i = ∑ (2.49) in the case of an EPSEM design.
- Stephen Gorard(Author)
- 2003(Publication Date)
- Continuum(Publisher)
Such an example emphasizes two things. First: despite its lack of popularity, judging from its rarity in the literature, random sampling is actually a lot easier than Stratified Sampling. Second: even where the purpose of the study is to collect new data (on the motivation of trainee teachers, for example), it is often important to conduct a fairly detailed secondary analysis first in order to identify the characteristics of the population being studied (see Chapter Two). 70 Quantitative Methods in Social Science Clustering Using a clustered sample implies not so much a difference in selection procedures as a difference in defining population units. The cases we are interested in often occur in natural clusters such as institutions. So we can redefine our population of interest to be the clusters (institutions) themselves and then select our sample from them using one of the above procedures. The institutions become the cases, rather than the individuals within them. This has several practical advantages. It is generally easier to obtain a list of clusters (employers, schools, voluntary organisations, hospitals, etc.) than it is to get a complete list of the people in them. If we use many of the individuals from each cluster in our selected sample, we can obtain results from many individuals with little time and travel, since they will be concentrated in fewer places. For example, in a survey of teachers we might select a random sample of 100 of the 25,000 schools in England and Wales, and then use the whole staff of teachers in each of these selected schools. As with systematic sampling, it is important that the odds of a cluster's being selected are in proportion to the number of individuals it represents (i.e. schools with more teachers should be more likely to be picked). Despite this complication in the calculation (and the need to have at least some information about each cluster), this approach is growing in popularity (see Chapter Eight).
Index pages curate the most relevant extracts from our library of academic textbooks. They’ve been created using an in-house natural language model (NLM), each adding context and meaning to key research topics.
