Virginia M. Lesser

Personalization of Mail Surveys for General Public and Populations with a Group Identity : Results from Nine Studies

Abstract The effect of personalization on mail survey response rates was examined in nine studies that included 17 comparisons under several research conditions. A study of this variable across multiple experiments in five agricultural experiment stations was undertaken because of conflicting results from previous research and from concern that the effectiveness of personalization might have decreased over time. Results show that, while response to general public surveys appeared to increase modestly across all treatment groups, there was no positive effect for populations in which a group identity (e.g., Dear Oregon Gardner or Dear ATV Owner) is employed to address respondents in cover letters. Personalization appears to remain useful for improving response in surveys of the general public.

Nonsampling Errors in Environmental Surveys

A number of probability surveys are in place in the United States to estimate and monitor changing conditions of ecological resources. The Environmental Protection Agencys Environmental Monitoring and Assessment Program (EMAP) is a more recently created program designed to estimate and monitor changing conditions of ecological resources. Nonsampling errors, as well as sampling errors, need to be investigated to assess total survey error. Specifically, the impact of frame, measurement, and nonresponse errors encountered when monitoring various ecological resources are discussed. Examples of nonsampling errors encountered within the EMAP program are examined. Nonsampling errors will also be examined for two more established probability surveys monitoring ecological resources, the U.S. Department of Agricultures (USDA) June Enumerative Survey and the USDAs National Resources Inventory. Finally, the remedies to adjust for nonsampling errors in these environmental surveys are discussed.

APPLYING SURVEY RESEARCH METHODS TO ACCOUNT FOR DENIED ACCESS TO RESEARCH SITES ON PRIVATE PROPERTY

Needs for environmental information have led to more emphasis on surveys that assess changing environmental conditions. In order to obtain biological, physical, and chemical measures of environmental change, it is critical to obtain repeated access to research sites. Use of established survey methodology can improve access rates to private property in order to visit randomly selected research sites. Surveys initiated by the U.S. Environmental Protection Agency’s Environmental Monitoring and Assessment Program (EMAP) were conducted in 1995 and 1996 in Eastern North Dakota to monitor environmental conditions in prairie wetlands. These surveys required written permission from landowners to access private property. A summary of survey research methods designed to improve access rates to sites on private property is presented. Approximately 20% of the landowners who granted access did so in response to telephone calls after attempts by mail failed. The data show that a disproportionate number of accessible sites were located on non-agricultural lands, particularly in 1996. Owners of sites where agricultural land abutted wetlands tended to deny access. This may introduce bias in the monitoring data. A weighting class adjustment procedure was applied to account for any nonresponse bias resulting from denied access. This procedure deals with the imbalance from denied access and at least partially returns a balance to the collected data.

A proposed methodology to determine accuracy of location and extent of geographically isolated wetlands

Wetlands can provide significant environmental benefits such as assimilation of pollutants, flood water storage, groundwater recharge, and fish and wildlife habitat. Geographically isolated wetlands (IWs) can provide the same benefits as other wetlands, but are particularly vulnerable to losses from urbanization and agriculture since they are isolated, often smaller, and not equal to other wetlands in regulatory protection. There is a need for a reliable assessment of the total number and area of IWs. Access to this information can provide the foundation to monitor IW loss, and can assist in understanding the ecological effects of such loss as well as help to address regulatory shortcomings. A probability sampling survey methodology that balances statistical considerations, expert opinion, and operational considerations is discussed for assessing the accuracy of IW maps. A simulation exercise was used to illustrate how the proposed sampling methodology will produce reliable results in map accuracy assessment. A wetland map for the Lockwood Folly River watershed, in Brunswick County, NC, was created using historical and recent data and used as the sampling frame in the simulation exercise. An updated wetland map for the same watershed, showing exact locations of wetlands, was used to provide “ground-truth” observations based on wetland delineations approved by the U.S. Army Corps of Engineers. Survey-based accuracy estimates were calculated by comparing site classification differences obtained by using both the original and updated wetland maps.

Risk behaviors and self-reported illnesses among Pacific Northwest surfers.

Although surfers have high incidental exposure to marine waters, no studies have investigated if surfer risk behaviors (such as surfing during advisories, near an outfall, during a rain event, or use of personal protective equipment) increase or decrease the risk of acquiring waterborne illnesses. We used a web-based survey to assess the association between risk-based behaviors and self-reported illnesses among Pacific Northwest surfers. Commonly reported illnesses include: ear infection or discharge (38%), sore throat or a cough (28%), diarrhea (16%), fever (10.5%), and vomiting (7%). Surfing often during rain events was associated with an increased likelihood of diarrhea (OR = 2.7; 95% CI: 1.4-5.47), sore throat (OR = 1.26; 95% CI: 1.01-2.05), and ear infection (OR = 1.39; 95% CI: 1.01-2.32). Surfing during a health advisory was associated with increased likelihood of diarrhea (OR = 1.94; 95% CI: 1.03-4.64) and sore throat (OR = 2.32; 95% CI: 1.23-4.40). Other behaviors associated with increased illnesses include body surfing, surfing near an outfall, frequency of surfing, and use of ear plugs. Approximately 40% of surfers were unaware if they had surfed during an active health advisory and 29% knowingly surfed during advisories, suggesting the need to engage this population about potential harm and behaviors that may increase health risk.

A comparison of periodic survey designs employing multi-stage sampling

Sampling designs considered for a national scale environmental monitoring programme are compared. Specifically, design strategies designed to monitor one aspect of this environmental programme, agro-ecosystem health, are assessed. Two types of panel survey designs are evaluated within the framework of two-stage sampling. Comparisons of these designs are discussed with regard to precision, cost, and other issues that need to be considered in planning long-term surveys. To compare precision, the underlying variance of a simple estimator of mean difference is derived for each of the two designs. A variance and cost model accounting for the different rotational sampling schemes across designs are developed. Optimum stage allocation for each design are assessed with the variance-cost models. The best choice of design varied with the conditions underlying the variance model and the degree of other sources of survey error expected in the programme.

A Permutation Test for Correlated Errors in Adjacent Questionnaire Items

Response patterns are of importance to survey researchers because of the insight they provide into the thought processes respondents use to answer survey questions. In this article we propose the use of structural equation modeling to examine response patterns and develop a permutation test to quantify the likelihood of observing a specific response pattern. Of interest is a response pattern where the response to the current item is conditioned on the respondents answer to the immediately preceding item. This pattern manifests itself in the error structure of the survey items by resulting in larger correlations of the errors for adjacent items than for nonadjacent items. We illustrate the proposed method using data from the 2002 Oregon Survey of Roads and Highways and report SAS code that can be easily modified to examine other response patterns of interest.

The National Children's Study 2014: Commentary on a Recent National Research Council/Institute of Medicine Report

The National Children’s Study 2014: Commentary on a Recent National Research Council/Institute of Medicine Report Marie C. McCormick, MD, ScD; Dean B. Baker, MD, MPH; Paul P. Biemer, PhD; Barbara Lepidus Carlson, MA; Ana V. Diez Roux, MD, MPH; Virginia M. Lesser, DrPH; Sara S. McLanahan, PhD; George R. Saade, MD; S. Lynne Stokes, PhD; Leonardo Trasande, MD, MPP; Greg J. Duncan, PhD From the Harvard School of Public Health, Boston, Mass (Dr McCormick); School of Medicine, University of California, Irvine, Calif (Dr Baker); RTI International, Research Triangle Park, NC (Dr Biemer); Mathematica Policy Inc, Boston, Mass (Ms Lepidus Carlson); Drexel University School of Public Health, Philadelphia, Pa (Dr Diez Roux); Oregon State University, Corvallis, Ore (Dr Lesser); Princeton University, Princeton, NJ (Dr McLanahan); University of Texas Medical Branch at Galveston, Galveston, Tex (Dr Saade); Southern Methodist University, Dallas, Tex (Dr Stokes); School of Medicine, New York University, New York, NY (Dr Trasande); and University of California, Irvine, Calif (Dr Duncan) Much of this commentary is drawn from NRC/IOM (2014). Conflict of Interest: The authors declare that they have no conflict of interest. Address correspondence to Marie C. McCormick, MD, ScD, Department of Social and Behavioral Sciences, Harvard School of Public Health, 677 Huntington Ave, Boston, MA 02115 (e-mail: mmccormi@hsph.harvard.edu). Received for publication August 22, 2014; accepted August 24, 2014. A CADEMIC P EDIATRICS 2014;14:545–546 T HE NATIONAL CHILDREN’S Study (NCS) was autho- NICHD specified that the panel’s review should cover such aspects of the Main Study design as the national prob- ability sample’s overall sample size and design; the use of hospitals and birthing centers as the primary sampling unit; the relative size of the prenatal and birth strata in the prob- ability sample; the size of the supplemental convenience samples; optimal use of sibling births; use of health care providers to refer prospective participants; the proposed study visit schedule, with emphasis on more frequent data collection in pregnancy and early childhood; the proposed approach to assess health and developmental phenotypes; and the proposed approach to define and char- acterize health disparities. The panel concluded that to meet its charge of evaluating the scientific merit of the Main Study, the NCS program office would need to provide specific documentation on the sampling design, the sample frame, the data collection protocols, and the study instruments, rather than just describe the conceptual framework, strategies, and antici- pated processes to design the Main Study. The panel re- viewed what the NCS program office provided by way of documentation and responses to several sets of written questions from the panel. The panel also listened to public comments during open meetings that included the NICHD staff. In addition, the panel engaged consultants to provide an expert cost analysis to support comparison of various design alternatives. rized by the Children’s Health Act of 2000 (Public Law 106-310) and is being implemented by a dedicated pro- gram office in the Eunice Kennedy Shriver National Insti- tute of Child Health and Human Development (NICHD). In the words of the program office, the NCS is planned to be a “longitudinal observational birth cohort study to eval- uate the effects of chronic and intermittent exposures on child health and human development in US children.” 1 The NCS would be the first US study to collect a broad range of environmental exposure measures for a national probability sample of about 100,000 children followed from birth or before birth to age 21. Detailed plans for the NCS were developed by 2007 and reviewed by an outside panel. 2 At that time, sample recruit- ment for the NCS Main Study was scheduled to begin in 2009 and to be completed within about 5 years. However, results from the initial 7 Vanguard Study (pilot) locations, which recruited sample cases in 2009–2010, indicated that the proposed household-based recruitment approach would be more costly and time-consuming than planned. In response, the program office conducted additional pilot tests in 2011–2013 to evaluate alternative recruitment methods. Based on these results, the study design was revised in early 2013, and a tentative 2015 start date was set for the Main Study. In March 2013, Congress requested a review of the revised study design by a panel of the National Research Council and Institute of Medicine. Congress stated the panel should “conduct a comprehensive review and issue a report regarding proposed methodologies for the NCS Main Study, including whether such methodologies are likely to produce scientifically sound results that are gener- alizable to the United States population and appropriate subpopulations.” 2 A CADEMIC P EDIATRICS Copyright a 2014 by Academic Pediatric Association O VERALL A SSESSMENT The National Research Council and Institute of Medi- cine released the panel’s report, The National Children’s Study 2014: An Assessment (NRC/IOM, 2014), on June 16, 2014. 3 The panel concluded that the National Chil- dren’s Study has the potential to add immeasurably to Volume 14, Number 6 November–December 2014

Conditional and unconditional estimators of population size

Estimators for the population sizes of animal species are similar to Horvitz-Thompson estimators — they involve dividing counts of detected animals by the probabilities of detection. Knowing detection probabilities for different subpopulations allows one to estimate each sub-population size with such an estimator, and to add the results for an estimator of the total population. In the case where the proportions of animals belonging to the different sub-populations are also known, this paper shows that using those proportions to arrive at a common average detection probability will, when used in conjunction with the total number of animals detected, result in a better estimator. We provide two examples where the inferior estimator may seem sensible.