China
Africa
Explore chapters and articles related to this topic
Sampling Theory
Published in Marcello Pagano, Kimberlee Gauvreau, Heather Mattie, Principles of Biostatistics, 2022
Marcello Pagano, Kimberlee Gauvreau, Heather Mattie
Formally, to select a stratified random sample, we first divide the population into H strata with the hth stratum having size Nh. We then treat each of these strata as a smaller population, make inference within each stratum, then combine across the strata. As an example, start with Table 21.1 which shows the estimated number of women between the ages of 15 and 49 in each of the 9 provinces in South Africa. From this population we obtain a stratified sample of size n by choosing a separate simple random sample of size nh from stratum h, for each of the H strata, such that
Standardizing an Assessment
Published in Lucy Jane Miller, Developing Norm-Referenced Standardized Tests, 2020
James Gyurke, Aurelio Prifitera
The initial step in securing a stratified random sample is to compose a list of all potential sources of the sample (i.e., for young children those sources might include day care centers, infant stimulation programs, preschools, church programs, YMCAs, rehabilitation centers, children’s hospitals, etc.). Using a random numbers table or other appropriate randomization method, the sources should then be randomly selected. From a list of all appropriate subjects at the sources, individual children can then be selected randomly. Permission to test the children should be obtained at this point, in addition to demographic information on each child pertaining to the designated stratification variables (i.e., sex, ethnic group, community size, parent’s educational level, etc.). An example of a biographic information form, to be completed by parents, is shown in Table 5. After reviewing these forms, testers can then determine which children should be tested to meet the quotas in each stratification cell. It is important to keep in mind that the more randomly the sources and the sample are selected, the greater the likelihood that resultant data will be error free and unbiased.
Evaluating Samples when Researchers Generalize
Published in Fred Pyrczak, Maria Tcherni-Buzzeo, Evaluating Research in Academic Journals, 2018
Fred Pyrczak, Maria Tcherni-Buzzeo
Comment: Students who are new to research methods are sometimes surprised to learn that there often is no simple answer to the question of how large a sample should be. First, it depends in part on how much error a researcher is willing to tolerate. For public opinion polls, a stratified random sample of about 1,500 produces a margin of error of about one to three percentage points. A sample size of 400 produces a margin of error of about four to six percentage points.28
Teachers’ Empathy for Bullying Victims, Understanding of Violence, and Likelihood of Intervention
Published in Journal of School Violence, 2022
Anett Wolgast, Saskia M. Fischer, Ludwig Bilz
The data set (from the research project “Teacher Action on Violence and Bullying” supported by the German Research Foundation) includes responses from n = 556 teachers (79.4% female) in 24 German schools. The teachers were 51 years of age on average (M = 50.6, SD = 8.4, Min = 27, Max = 65) and had an average of 26 years’ teaching experience (M = 26.3, SD = 10.1). The teachers worked in seven academic-track schools (Gymnasium), 13 vocational-track schools (Realschule) and four special education schools. The distribution of teachers across the different school types in the sample corresponds to the distribution among all teachers in the federal state in which the study was conducted (Bilz et al., 2016). The drawing of the stratified random sample (stratification characteristic: school type) was carried out using probability-proportional-to-size design (Skinner, 2016). To obtain the 24 participating schools, 40 schools were contacted (participation rate at school level: 60% and at teacher level: 59%).
The economic burden of uncontrolled gout: how controlling gout reduces cost
Published in Journal of Medical Economics, 2019
Natalia M. Flores, Javier Nuevo, Alyssa B. Klein, Scott Baumgartner, Robert Morlock
This project includes data from the 2012 and 2013 US NHWS (2012 NHWS: n = 71,157; 2013 NHWS: n = 75,000). Respondents who took the survey more than once during the 2-year period were only included once and their most recent responses were used, which resulted in n = 130,089 unique respondents. The NHWS is a self-administered, Internet-based questionnaire from a sample of adults (aged 18 or older). A stratified random sample (with strata by gender, age, and ethnicity) was implemented to ensure that the demographic composition of the sample is aligned to that of the corresponding adult population as measured by the US Census. Several previous publications have favorably compared the NHWS, and some of its prevalence estimates, with other governmental sources16–18. The NHWS has received approval from Essex Institutional Review Board (IRB) (Lebanon, NJ).
An application of nonparametric regression to missing data in large market surveys
Published in Journal of Applied Statistics, 2018
Gary Madden, Nicholas Apergis, Paul Rappoport, Aniruddha Banerjee
The survey is a stratified random sample with stratification based on the key demographics: gender, age, household income, race/ethnicity, region of country (balanced approximately to national proportions). Operationally, households are drawn from Internet panel maintained by Luth Research, Inc., San Diego, CA. Questionnaire responses are provided by 4743 individuals (includes household heads and other eligible members of their households; with only respondents 19 years and over surveyed) residing in 2584 households. The sample for estimation is comprised of 4560 individual survey responses.2