Home > Free Essays > Sociology > Everyday Interactions > Sampling and Survey Design Pre 2001

Sampling and Survey Design Pre 2001 Research Paper

Exclusively available on IvyPanda Available only on IvyPanda
Updated: Sep 9th, 2021


Recently a lot of changes have occurred that require our reassessing the compilation of topics traditionally trained in a survey sampling route and the methods we use to educate such a course. For instance, recent investigate in sampling has listening carefully on nonsampling errors, small-area estimation, and model-based methods. In adding, the wide ease of use of public-use information from large-scale example surveys, and the processor software to examine such data, income that there are now many minor data users who need to study sufficient to analyze the information suitably.

With these issues in brain we held a panel conversation on education survey sampling at the August 1995 Joint Statistical Meetings in Orlando, FL. The board members transport a broad range of knowledge and viewpoints to the conversation. Several panelists contain knowledge operational as survey statisticians in administration agencies. They have also trained survey sampling, mainly to nontraditional scholars employed in government. The additional panelists are academicians with knowledge education survey sampling, writing sampling texts and software, and in arithmetical teaching (Ronald S. Fecso, 1996).

In the board session these experts discussed how they believe we should be education sampling today, what topics be supposed to be covered, and what be supposed to be fall from the traditional course. They described what they do at what time they teach sampling, how the student responded to the course fabric and what seems to work best for serving students to understand the issue related to sampling.

Although the board discussion was listed in the penultimate session of the combined Statistical Meetings, it was well be present at and was followed by an energetic discussion and pleas for additional advice on education survey sampling. This article is an effort to help reply those pleas. In the following we there an edited account of the remarks from the board conversation.

As members of a board on education survey sampling we must consider how to contract with the “diversity” of subjects mentioned in the foreword. My comments will center on how diversity issues tell to teaching sampling lessons. For example, is the variety of students’ backgrounds a power or weakness in survey courses? Likewise, do the textbooks characteristically second-hand in survey courses labor with a varied class?

In the near the beginning 1980s, when I trained my first sampling route at the USDA Graduate School, the majority students had scholar mathematics majors, and a lot of had completed a mark off statistics degree or at smallest amount some statistics route work. Thus taking a semester or two accommodations to cover the majority of the material in a book similar to Cochran’s Sampling Techniques (1977) worked sensibly well. This arrangement was not too different from so as to in the sampling classes initiate at the USDA Graduate School in excess of 40 years ago by Dr. Deming (Linda L. Stinson, 1999).

USDA Graduate School students, usually working professionals, are rather different inhabitants from that establish in institution of higher education lessons. Yet, there have been alike change in surplus of time in the backgrounds of persons students. As the 1990s approached variety increased. More new students in the sampling class move violently with the material connected to sampling theory since their mathematical backgrounds were additional incomplete.

This change in the level of arithmetical preparedness comes with a alter in the knowledge and backdrop of the students deficient to study survey sampling.

Classes in the history consisted more often than not of government-employed statisticians operational on surveys. Now workers of contract survey firms and manufacturing, and individuals seem for a career change, often create up half of the students. Although these student have considerable information of survey operations acquired on the job, the instructive backdrop of the nongovernmental employees characteristically does not comprise a math main. Yet these students were frequently hired to be “the sampling being” in an association that lacks knowledgeable “samplers,” thus denying knowledge opportunities that are frequently obtainable in the larger arithmetical agencies. The backdrop differences complete the traditional fabric difficult to there. To accommodate the branch out student inhabitants we tear the two-quarter theory route into a noncalculus prerequisite route and a calculus-level route. Later we additional special themes courses (e.g., advanced topics in sampling hypothesis and topics in organization surveys) to give coverage of significant material left exposed in the shortened arrangement.

Impact Diversity In Educational Background Has On A Sampling

If we analyzed then we come to know that to illustrate the impact variety in educational backdrop has on a sampling course I will employ a familiar model alike to the USDA food pyramid. Figure 1 shows the survey pyramid. I am not in conflict that this graphic present an exhaustive catalog of survey tasks or the most excellent terminology for the survey everyday jobs, but the pyramid does reproduce how the quantity of survey work is covered according to the quantity of “human assets” expended. Typically, most time is exhausted obtaining reply – the time spent by populace responding to the survey and by populace concerned in data compilation behavior as enumerators or phone interviewers. Data editing consumes a marvelous quantity of effort in a lot of organizations, though there is a group toward risk-based negligible restriction approaches. The everyday jobs in a survey from information collection up from side to side and counting design issues use less and fewer people in a typical association. The pyramid’s peak stands for the few populaces in each organization operational on technological issues such as discrepancy estimation. With this organizational replica for sampling work we may look at the association between the sampling route content and these tasks. Specially, what survey knowledge is most wanted for populace to achieve these tasks efficiently?

In a typical route taught 10 or 15 years ago the majority of the class time was exhausted at the climax of the pyramid, focusing on plan and their various discrepancy estimates. Very little time was exhausted on issues connected to the ready tasks found close to the bottom of the pyramid. It seems clear from the turnout at this session that the old course outline does not labor well and wants “reengineering.” I argue that quite than lament the termination of the old route, we be supposed to be pleased with our new chance. We now have more populace who are paying attention in survey issues, particularly those issues connected with tasks found close to the pyramid’s base. Although we do require more people trained to employment on the technological problems linked with the climax of the pyramid, we require far more people for the additional jobs. The customary course does not get ready students for person’s jobs.

This panel’s center is on an environment in which present is only one chance to have a scholar in a survey course. What do we desire to do in that course? Should we center on the pinnacle of the pyramid in such a course? How do we discover the time to wrap other material?

We might start reengineering the on one occasion sampling course by investigative the needs of boss. For example, though the report of the Subcommittee on Measurement of Quality in Establishment Surveys (1988) is a bit behind the times, it is probable still deep of the skills needed in a characteristic administration establishment survey. The subcommittee accounts that “25 percent of the surveys appraisals had no variance estimates and 45 percent didn’t issue discrepancy estimates, whilst 70 percent had no cognitive learning done, 35 percent didn’t problem to do any pretest of opinion poll, and 75 percent didn’t have post-enumerative learns.” Thus the difficulty is not just that variances were not life form computed generally; other aspects of the survey procedure that are essential are absent or apparently underutilized in numerous of the organizations deliberate. I doubt that the universal population of survey associations fares much improved and some strength argue that it might be significantly not as good as.

To speak to how the fabric taught in a sampling route relates to service in a survey organization I will focus on the central administration statistician. First, a small number of words concerning what the name “government statistician” means may be helpful. The pyramid is old to portray the different tasks complete by administration statisticians in central survey associations. There are two classifications of government statisticians: the arithmetical statistician (called the 1529 series) and the survey or universal statistician (called the 1530 series). Most of the populace with the title of administration statistician is in the 1530 series. Entry supplies for the 1530 sequence are not difficult: 6 hours of statistics and 9 hours of arithmetic or figures, for totals of 15 hours to be eligible. Some might argue that it is too simple to “get in” with the name of statistician. These employees, though, are in charge of the volume of the survey work, treatment to the behavior that forms the bottom of the pyramid. They labor on the day-to-day tasks of gets and dispensation the information (James E. Grizzle, 1995).

According to the expert analysis some sight these service “requirements” as too feeble. What must be documented is that these are admission requirements, and not the labor force normal. Entry-level salaries by means of such a background are in the near to the ground $20,000s. You will have a hard time finding anybody with survey methodology preparation eager to work for $20,000. In fact, it is often hard to find people with the smallest quantity supplies who are eager to work for that pay. The basic obligation gives government organizations the aptitude to hire people who, if administers and trained correctly, can go forward to places requiring information of more advanced method. Meanwhile, we require these people to take out the size of the survey procedure (Alain Theberge, 1999).

Because we need a variety of training experiences for the variety of survey tasks, come again? do we want in a solitary survey course? Table 1 lists a figure of idea that a scholar be supposed to learn. Other panelists will address the specific concepts, so I will make but one comment. Today we also require having more width in the one-time course. We require wrapping more than design potential, estimators, and discrepancy computation. With this require over 50 years old this sitting is long in arrears (Michael A. Gips, 2005).

There are too a lot of populace in other job sequence, such as the computer-related sequence, working on reviews. Given the nature of central statistical labor and of the agencies utilizes the majority of the statisticians, surely the bulk of the statistician’s labor on surveys at least some time throughout their administration careers. Many of them do survey agreement monitoring, and do not require a predominately arithmetical exposure to surveys, but they require contact to the width of survey problems and possible approaches to answers.

In preparing for the board discussion I complete several contacts by means of other government organization. Although this was a extremely little and nonrandom example, the respondents were from organization use a large labor force of statisticians. My organization contacts point to that they want populace to understand and be grateful for survey troubles, and not just study the “formula in the box.” They would like to see container studies in a route. They particularly like the idea of a workshop, and they would like to see novel hires pending in with information of sampling-related software. This wish for students with hands-on knowledge is not original. For example, Jensen (1948) discusses an near the beginning plan to provide internships for scholars to study sampling and labor at a government organization (Occupational Hazard, 2002).

This research focused on this truth that what should we do when altering this one course? It is not just a problem with obtainable textbooks. Referring to the pyramid, present are textbooks that might be used for every single of the layers. The difficulty is that if you have scholars for only one semester, you cannot go from side to side all the textbooks, and the preponderance of the schoolbooks is far too multifaceted for a first-time scholar anyhow.

Sampling Theory

The subject we often tackle is that the route is labeled “sampling hypothesis,” but we find that for a lot of students there actually is not much hypothesis we can do. I do not believe it is bad to have your one route geared in the direction of sampling theory if the majority of your scholars have homogeneous arithmetical backgrounds and can grip the fabric. Otherwise, having a “sampling-theory” route and watering the fabric down to meet your scholar mix is not responsibility anyone a good turn (Martina Morris, 2004).

What strength we do if we have single course? The Computer discipline Department strength offer a route on database organization related to survey application; the Business School might focus on survey procedure management. Then scholar could take more than a few of person’s courses throughout their time at the organization. Maybe parts of the course might be trained jointly so everybody gets some core concept. Students would then labor on additional parts of the width of sampling as it relates to the section offering the course. Have a survey behavior by the combined lessons, where each group of students works as a component in a survey association; one collection works on the questionnaire, one more works on the computer designs, one more does the restriction and maybe scholars from the School of Business could administer the venture. You could randomize meeting to students and lessons so that they might seem at interviewer effects, the type of preparation received, plus its effect on interviewing. I be acquainted with so as to is a tall order. Maybe what we need is some funding beginning a local manufacturing, government, or from the organization – someone deficient a survey done. A financial inducement might inspire different departments to expand sampling courses jointly.

In raising our new course(s) we must dedicate considerable time to ask for scholar feedback in mounting course resources and computer agendas. We need to study come again? Texts work with this move toward, and we require projects that labor.

A network can be distinct as a set of persons and the relations among them. The combined focus on people and partnerships is accountable for all of the dissimilarity flanked by a network survey and an additional traditional survey. It has an crash on sampling, tool design, and psychoanalysis. Each is described underneath.

Network Survey And Sampling

Collecting information from a network is alike to collecting information as of a population of people in one intelligence there is a variety of sampling strategies that stretches from a poll, where the inhabitants and the sample are the same (sometimes called a diffusion sample), to a simple chance sample that is destined to be representative of the inhabitants. The latter is radically less expensive, but requires hold up from arithmetical sampling hypothesis to be envoy. There is a wide variety of sampling theory obtainable for studies in which the person (be it a person, a household, a firm, etc.) is the unit of psychoanalysis. If we want to be acquainted with what amount of the inhabitants has been vaccinates against polio, or what amount of households have at smallest amount one child beneath 6 years old in house, there are a lot of ways to design an efficient likelihood sample. This is not to speak that the sensible completion of a true likelihood sample is easy or cheap, but we be acquainted with what has to be done. When the thing of interest is a system structure, however, very small sampling theory is available. Say we required to know whether a system was totally connected, or how frequently a person was probable to be linked (directly or indirectly) to ten or additional other people, or how often partnerships concerned persons by means of more than a 5-year age dissimilarity. We know a great deal less concerning how to design a likelihood sample for investigate query like this. The difficulty is that we rarely be acquainted with how to set up the likelihood that any exacting unit (person, partnership, or system component) was chosen into the example (the “addition probability”). Without this, there can be no deduction from the example we have to the inhabitants of attention.

Difference between A Network Study and More Traditional Behavioral

If we analyzed then we come to know that the key dissimilarity between a system study and more traditional behavioral studies is that information are needed on the relations as well as the people. This means there are two dissimilar sampling units: the person respondents, and the company. Individuals are motionless the source of all in order, of course, but the in order they provide is not incomplete to their own quality it includes information that will help set up the pattern of relations between them and everybody else. The sampling border for these two units is consequently nested: first we choose how to example respondents, and then we decide how to example company from these respondents. The essential variations in system study designs are resulting from the habits in which these two heights are sampled.

The range of network sampling design is distinct by the type of example second-hand for choose respondents. At one end of the range is a “saturation sample, ” or poll of the pertinent inhabitants, and we use the term complete network design to explain this move toward. Other conditions in the writing include socio-metric and worldwide system designs. There is no difficulty of deduction here, but the cost is that we have to enroll the entire inhabitants. At the other end of the range is the simple random example of respondents, and we use the word local network design to explain this approach. One more common term in the text is selfish design. Here the addition probability of respondents is recognized, but one can expect that there determination be much less in order on the network arrangement formed by the relations in the inhabitants.

Survey and Sampling Design

Of the three designs, the incomplete network design is the majority multifaceted. A broad range of strategies is enclosed by this caption since after the initial example is enrolled, present is extra sampling involved at two levels: partners and generations. We can register some somewhat than all of the chosen partners at each age group, and we can trace out a number of quite than all generations in the inhabitants. Enrolling all associates, and all generations give small control over example size. The classic increase sample is a incomplete network plan. In this design, all associates of the first respondent are enrolled, for a set figure of generations, characteristically set a priori. With a number of information on the characteristic size of a local system, this design gives a number of controls over final example size, but we be acquainted with little concerning the resulting addition probabilities. Uninfected persons are thus endpoints, and the sampling stops at what time the last set of associates are all uninfected (in hypothesis at least). This approach is very alike to the “adaptive samples” that are now old in the physical sciences to obtain well-organized estimates of the size of little populations. It is likely to set up the inclusion likelihood for each being in a true adaptive example, but this approach once more provided little control in excess of final example dimension.

Designs Level Of Local, Partial, and Complete

This research focused on this truth that these three designs local, incomplete, and total—represent the range of strategies obtainable at the level of person enrollment, and the input differences in system survey design. To a large degree, the portion of the system that is exposed by the example is strong-minded by which of the respondent sampling plan is selected.

There is one more level of sampling that have to be done to recognize the links, though, and the way that this is complete has dissimilar implications for every design.

The fact that we are by means of a next form of sampling to recognize links (or partnerships) is frequently not appreciated. But a system is distinct by the links as a great deal as the nodes. Even if we have a diffusion sample of persons, as we do in the absolute system design, our in order on the partnerships is incomplete to what the respondents inform us, and that depends on what we have asked them. This may be the cause that the sampling of relations is often not openly recognized as fraction of the network sampling procedure: it happens in the survey. The nesting of the two sampling units, persons then partnerships, means that fraction of the sampling happen in the background of scheming the survey gadget.

Survey Instrument Design

According to the expert analysis the method we use to sample the companies in the questionnaire is called a “name producer” the query that is used to elicit and recognize associates. It may or may not in fact generate the real names of partners, but it does draw out a set of exclusively identified partners, not now a numeric total. This is come again? Distinguishes a true system instrument from a normal survey instrument. Characteristic name producers in the context of HIV investigate strength is “Thinking about the being (or persons) you have had sex with in the history six months, and catalog their initials at this time.”

If we analyzed then we come to know that there are two parts to each name producer: the family member, and the sampling constraint. The family member in the example on top of is “sex partner.” The sampling restraint is the last 6 months. There are a lot of variations in every, and even as there is no complete typology that defines these differences, there are a number of useful strategy.

Name generators are fraction of the survey, but they are a sampling device in network studies. For the restricted and total network designs, they give the example of partnerships that will be analyzed. For incomplete network designs, they in addition affect the example of respondents due to the connection tracing operational in these designs. Name generators in this study determine the list of people eligible to be example in each subsequent age group. If we picture the network survey as a flashlight that excel on certain parts of the system, creation them visible, the mixture of respondent staffing and name manufacturer are that flashlight.


Article Title: Teaching Survey Sampling. Contributors: Ronald S. Fecso – author, William D. Kalsbeek – author, Sharon L. Lohr – author, Richard L. Scheaffer – author, Fritz J. Scheuren – author, Elizabeth A. Stasny – author. Journal Title: The American Statistician. Volume: 50. Issue: 4. Publication Year: 1996. Page Number: 328+. COPYRIGHT 1996 American Statistical Association; COPYRIGHT 2002 Gale Group.

Teaching survey sampling by Ronald S. Fecso , William D. Kalsbeek , Sharon L. Lohr , Richard L. Scheaffer , Fritz J. Scheuren , Elizabeth A. Stasny

Recently many changes have occurred that necessitate our reassessing the collection of topics traditionally taught in a survey sampling course and the methods we use to teach such a course. For example, recent research in sampling has focused on nonsampling errors, small-area estimation, and model-based methods. In addition, the wide availability of public-use data from large-scale sample surveys, and the computer software to analyze such data, means that there are now many secondary data users who need to learn enough to analyze the data appropriately.

With these issues in mind we held a panel discussion on teaching survey sampling at the August 1995 Joint Statistical Meetings in Orlando, FL. The panel members brought a wide range of experience and viewpoints to the discussion. Several panelists have experience working as survey statisticians in government agencies. They have also taught survey sampling, mainly to nontraditional students employed in government. The other panelists are academicians with experience teaching survey sampling, writing sampling texts and software, and in statistical education.

In the panel session these experts discussed how they think we should be teaching sampling today, what topics should be covered, and what should be dropped from the traditional course. They described what they do when they teach sampling, how the students responded to the course material, and what seems to work best for helping students to understand the issues related to sampling.

Although the panel discussion was scheduled in the penultimate session of the Joint Statistical Meetings, it was well attended and was followed by a lively discussion and pleas for more advice on teaching survey sampling. This article is an attempt to help answer those pleas. In the following we present an edited version of the remarks from the panel discussion.

Article Title: Measuring How People Spend Their Time: A Time-Use Survey Design. Contributors: Linda L. Stinson – author. Journal Title: Monthly Labor Review. Volume: 122. Issue: 8. Publication Year: 1999. Page Number: 12. COPYRIGHT 1999 U.S. Bureau of Labor Statistics; COPYRIGHT 2002 Gale Group.

Measuring how people spend their time: a time-use survey design. by Linda L. Stinson.

Methodological decisions concerning the mode, follow-up probes, coding schemes, and simultaneous activities all have far-reaching implications

Time-use studies typically have a single focus: to study the frequency and duration of human activities. For example, time-use surveys may ask respondents to report everything they did during a 24-hour period along with some indication of the starting and stopping times of those actions. This chronological reporting procedure avoids many pitfalls that other survey estimation procedures encounter and is less subject to distortion due to “social desirability bias.” But there are many methodological considerations to take into account when designing a time-use survey. Decisions concerning reporting procedures and mode of data collection may influence data quality. Likewise, the choice of follow-up probes and the treatment of simultaneous activities can determine the amount of information available for accurate and reliable coding of activities.

This article describes the methodological decisions that the BLS time-use working group faced when designing a possible time-use survey.(1) It also presents the methodological choices that the group made and provides the rationale for those selections.

While time-use research (that is, the actual enumeration of the activities people perform) may have originated within the social sciences and the time management domain of the business world, international governments have also been quick to recognize the value of this information. A number of their policy-related issues can be addressed with time-use data. Consequently, the question is not so much “why?” time-use data should be collected, but rather, “how?” it can be collected well.

Article Title: A Split Questionnaire Survey Design. Contributors: Trivellore E. Raghunathan – author, James E. Grizzle – author. Journal Title: Journal of the American Statistical Association. Volume: 90. Issue: 429. Publication Year: 1995. Page Number: 54+. COPYRIGHT 1995 American Statistical Association; COPYRIGHT 2004 Gale Group.

A Split Questionnaire Survey Design. by Trivellore E. Raghunathan , James E. Grizzle.

The design of surveys must balance many competing goals. Due to the financial burden of recruiting or selecting individuals for studies, many survey questionnaires are formed by pooling questions from existing survey instruments in an attempt to obtain information that can be used for a variety of purposes. An example that motivated the work presented in this article is the Cancer Risk Behavior Survey conducted periodically by the Surveillance Investigation Section of the Cancer Prevention Research Program of the Fred Hutchinson Cancer Research Center. The survey, a probability sample of households in the state of Washington, interviews a male and a female age 18 years or older from each household by telephone. The sample is selected using random-digit dialing with an oversampling of households from large cities or communities. The objective of the survey is to estimate the prevalence of risk behaviors, such as smoking, alcohol consumption, and dietary factors, and the percentage of people who periodically undergo preventive care, such as mammography and Pap smear. It also asks about knowledge and attitude about various risk factors and their association with cancer and heart disease. A typical interview time is about 30 minutes, although it can take as long as 50 minutes.

Pooling questions to make a long questionnaire, although cost-effective, increases the response burden on the individuals sampled and may result in a large nonresponse rate. Several studies (Adams and Darwin 1982, Dillman, Sinclair, and Clark 1993, and Roszkowski and Bean 1990, to mention only a few in the vast literature) show empirical evidence that surveys with long questionnaires tend to have high non-response rates. Herzog and Bachman (1980) suggested that the quality of response may also be affected with the increasing length of the questionnaire.

The response rate in the Cancer Risk Behavior Survey has been around 70%, with more item nonresponse during the later pan of the questionnaire. Also, in 26% of the cases the second member of the household refused to respond after finding out how long the first member of the household took to complete the survey. A concern in the Cancer Risk Behavior Survey is that busier and healthier individuals may be unit nonresponders and that sicker individuals might more often be item nonresponders. Thus a question of interest is: What modification can be made at the design stage to reduce the responder burden and possibly increase the response rate?

A solution is to split the long questionnaire into several components and administer only a fraction of these components to any sampled individual. The number of components to be administered can be determined based on the survey objectives. For example, if the inferential quantities of interest are the population means, variances, and covariances, then one needs to elicit responses on every pair of variables from a group of individuals. One can accomplish this by administering only two components to any sampled individual such that every pair of components is administered to a few sampled units.

Article Title: Extensions of Calibration Estimators in Survey Sampling. Contributors: Alain Theberge – author. Journal Title: Journal of the American Statistical Association. Volume: 94. Issue: 446. Publication Year: 1999. Page Number: 635. COPYRIGHT 1999 American Statistical Association; COPYRIGHT 2004 Gale Group.

Extensions of Calibration Estimators in Survey Sampling. by Alain Theberge.

The method of calibration for estimation of population totals was described by Deville and Sarndal (1992). An implicit objective of the method is to use auxiliary information to obtain estimators that are approximately unbiased with a variance smaller than that of the Horvitz-Thompson estimator. The weights of sampled units obtained through calibration are as close as possible to the Horvitz-Thompson weights, while satisfying a certain constraint called the calibratiOn equation. When a weighted sum of squares distance is used to measure the distance between the two sets of weights, the estimator obtained through calibration corresponds to a generalized regression estimator. For more general distance measures, under certain regularity conditions, the estimator obtained can be approximated by a generalized regression estimator (see Deville and Sarndal 1992). Estimation software based on the method of calibration, or that of generalized regression, has been developed (see, e.g., Estevao, Hidiroglou, and Sarndal 1995).

Book Title: Network Epidemiology: A Handbook for Survey Design and Data Collection. Contributors: Martina Morris – editor. Publisher: Oxford University Press. Place of Publication: Oxford. Publication Year: 2004. Page Number: iii.

Overview of Network Survey Designs BY martina morris.

A network can be defined as a set of persons and the relationships among them. The joint focus on persons and partnerships is responsible for all of the differences between a network survey and a more traditional survey. It has an impact on sampling, instrument design, and analysis. Each is described below.

Collecting data from a network is similar to collecting data from a population of persons in one sense—there is a range of sampling strategies that stretches from a census, where the population and the sample are identical (sometimes called a saturation sample), to a simple random sample that is meant to be representative of the population. The latter is dramatically less expensive, but requires support from statistical sampling theory to be representative.

Article Title: Survey: 3.3 Million Workers Use Respirators. Magazine Title: Occupational Hazards. Volume: 64. Issue: 5. Publication Date: May 2002. Page Number: 25+. COPYRIGHT 2002 Penton Media, Inc.; COPYRIGHT 2004 Gale Group.

Survey: 3.3 Million Workers Use Respirators.

Results of a survey conducted in late 2001 by the Bureau of Labor Statistics (BLS) found that in the previous year, some 3.3 million employees used respirators, but only just over half of those workers knew why they were wearing the respirator or were taught how to use it properly.

The National Institute for Occupational Safety and Health (NIOSH) and other agencies will use the results from “Respirator Use and Practices” to develop new recommendations for the use of respirators in workplaces and design new research projects to improve the effectiveness of respirators.

A sampling of survey results:

  • In the 12 months prior to the survey, respirators were used by 3.3 million employees in 281,800 U.S. workplaces, which represent 3.1 percent of all employees and 4.5 percent of all workplaces.
  • Dust masks were the most commonly used respirator, comprising 71 percent of all respirators used. Seventeen percent of workplaces reported using air-supplied respirators in the past year.

Article Title: Student Survey: A View from the Trenches of Training. Contributors: Stephen Nunns – author. Magazine Title: American Theatre. Volume: 14. Issue: 1. Publication Date: January 1997. Page Number: 38+. COPYRIGHT 1997 Theatre Communications Group; COPYRIGHT 2002 Gale Group.

Student survey: a view from the trenches of training by Stephen Nunns.

The “Approaches to Theatre Training” issue survey can be one of the more daunting aspects of the editorial process at American Theatre. Yet despite its vastness and the challenges inherent in its compilation, the survey provides a notable opportunity – a moment of reflection when we can take the pulse of the eclectic and sometimes confusing world of theatre education.

In the past two years, the magazine has solicited responses from an array of educators, asking them questions about the changing nature of training in this country, due to pragmatic matters (like giving students the sort of practical skills that will make them employable after their education is done) and the outside pressures of what the 1996 survey called “our media-tized society.” Both earlier surveys have made two assumptions: that the teachers are the ones with the answers; and that the basic tenets of theatre training must change in some fundamental way in our new high-speed consumer culture.

This year we began by immediately discarding one of these assumptions – and turning to the students in American theatre-training programs for their in-the-trenches points of view. We sent the following question to schools of every size and shape in just about every comer of the country:

Why have you made a commitment to training for the theatre at this time when work in the other arts and entertainment media seems to be so much more profitable and to provide so much more visibility? As you fashion your career, how do you anticipate that working in the theatre will fit into it?

This research paper on Sampling and Survey Design Pre 2001 was written and submitted by your fellow student. You are free to use it for research and reference purposes in order to write your own paper; however, you must cite it accordingly.
Removal Request
If you are the copyright owner of this paper and no longer wish to have your work published on IvyPanda.
Request the removal

Need a custom Research Paper sample written from scratch by
professional specifically for you?

801 certified writers online

Cite This paper
Select a referencing style:


IvyPanda. (2021, September 9). Sampling and Survey Design Pre 2001. https://ivypanda.com/essays/sampling-and-survey-design-pre-2001/


IvyPanda. (2021, September 9). Sampling and Survey Design Pre 2001. Retrieved from https://ivypanda.com/essays/sampling-and-survey-design-pre-2001/

Work Cited

"Sampling and Survey Design Pre 2001." IvyPanda, 9 Sept. 2021, ivypanda.com/essays/sampling-and-survey-design-pre-2001/.

1. IvyPanda. "Sampling and Survey Design Pre 2001." September 9, 2021. https://ivypanda.com/essays/sampling-and-survey-design-pre-2001/.


IvyPanda. "Sampling and Survey Design Pre 2001." September 9, 2021. https://ivypanda.com/essays/sampling-and-survey-design-pre-2001/.


IvyPanda. 2021. "Sampling and Survey Design Pre 2001." September 9, 2021. https://ivypanda.com/essays/sampling-and-survey-design-pre-2001/.


IvyPanda. (2021) 'Sampling and Survey Design Pre 2001'. 9 September.

Powered by CiteTotal, free citation creator
More related papers