Research

General Research Areas
Current Research Projects

• National Resources Inventory
• Small area estimation for Census data
• Statistical models for estimating prevalence of food insecurity
• Welfare survey design
• Geospatial data in mobile field collection
• Rathbun Lake Watershed Health Assessment
• Nonparametric model-assisted estimation for forest resources

Past Research Projects

• Estimation procedures for soil map units
• Estimating usual dietary intake distributions
• Nitrogen runoff prediction

General Research Areas in CSSM

Sample designs: Combining samples and frames for interagency inventories, multi-phase designs, partial rotation designs, spatial sampling, dual frame designs

Survey estimation: Distribution estimation, imputation, multi-phase estimation, nonresponse and measurement error models, regression estimation, replication variance estimation, small area models, time series and longitudinal analysis

Survey methods: Mobile data collection systems, integrated computer-assisted tools for survey process, digital geospatial data in data collection

Current Collaborative Research Projects

National Resources Inventory

As part of a long-standing USDA Natural Resources Conservation Service (NRCS) cooperative agreement established in 1956, CSSM faculty and graduate students investigate statistical methods for natural resource inventory and monitoring surveys, with particular emphasis on national land use, soil erosion, and agro-environmental surveys. Current research interests include multi-phase sample designs for special studies, partial rotation designs for converting the current five-year panel sample for the National Resources Inventory to an annual data collection cycle, imputation methods for incorporating census and area segment data in point level data, fractional imputation for data missing by design, small area estimation, replication variance estimation, and computer-assisted methods for survey data collection.

Small area estimation for Census data

Wayne Fuller and other members of CSSM cooperate with staff of the Census Bureau on time series estimation methods and survey methodology. An estimation procedure closely related to regression estimation is being developed for possible application in the weighting of long form Census data. Research on small area estimation was conducted analyzing data from the 1990 Census.

Nonparametric model-assisted estimation for forest resources

Jean Opsomer is collaborating with Jay Breidt (Colorado State University) and Gretchen Moisen (Rocky Mountain Research Station, US Forest Service) on developing nonparametric and semi-parametric model-assisted estimators applicable to surveys of forestry and other natural resources. These models are expected to be useful for multi-phase surveys with remotely sensed auxiliary information, longitudinal surveys with partially overlapping rotation samples, and surveys with item nonresponse.

Geospatial data in mobile field data collection

Sarah Nusser and Les Miller (Computer Science) are investigating new frameworks for using and collecting geospatial data in mobile field computing environments, in collaboration with Mike Goodchild and Keith Clarke in the National Center for Geographic Information and Analysis at the University of California, Santa Barbara. Developments in mobile location-aware technologies and distributed geospatial databases are being harnessed to transform the use of geospatial information in field data collection as a reference resource, as auxiliary data in selecting sample locations, and in collecting spatial information from a field site. Research is being conducted on flexible infrastructure designs that will support ad hoc and prepared queries for spatial data generated from small field computing devices under limited bandwidth conditions. Software tools for manipulating spatial data and selecting samples are being developed, and extensions from current handheld technologies to wearable computing and augmented reality environments are being pursued. The research is supported by NSF and several federal statistical agencies, including the Census Bureau, USDA, USGS, and Bureau of Labor Statistics. http://dg.statlab.iastate.edu/dg

In a related NSF project, Nusser, Miller, Taps Maiti, and Hal Stern (Statistics) are working with researchers at the San Diego Supercomputer Center and the University of Maine to develop quality-aware query processing methods. The ISU team is examining statistical descriptors of geospatial data quality that can be used as decision parameters in smart query processing protocols and to combine geospatial data to create a query result.

Past Collaborative Research Projects

Statistical models for estimating prevalence of food insecurity

Jean Opsomer and Sarah Nusser participated in a research project involving a cooperative agreement with the USDA Economic Research Service to investigate the underlying theory and application of item response theory to the problem of estimating the prevalence of food insecurity in the U.S., and to develop shortened question sets for rapid assessment of food insecurity. This was joint research with Helen Jensen (Economics and CARD) and Amy Froehlich (Statistics).

Welfare survey design

A project sponsored by the Department of Health and Human Services and the USDA Economic Research Service was established to investigate methods of conducting state-level welfare surveys in a manner that allow the state-level data to be combined with data from national sources, such as the Survey of Program Dynamics (SPD). The project used a dual frame approach, using administrative welfare participant databases to augment a traditional area or Random Digit Dialing (RDD) sampling frame. Questions of local relevance were integrated with questions from the SPD questionnaire. Sarah Nusser; Cynthia Needles Fletcher, Steven Garasky (Dept. of Human Development and Family Studies); and Helen Jensen (Dept. of Economics) were co-investigators on this project.

Rathbun Lake Watershed Health Assessment

Jean Opsomer collaborated with Richard Cruse (Agronomy), Thomas Isenhart (Forestry) and researchers at Chariton Valley Resource and Conservation on a comprehensive study of water quality impairments in the watershed. As part of the study, erosion from all sources was quantified on a random sample of plots, and small area estimation was used to predict the erosion at the sub-watershed level.

Estimation procedures for soil map units

As part of a second cooperative agreement with the NRCS, national soils databases were maintained at ISU since 1972. A collaborative project with the ISU Agronomy Department (Tom Fenton, Dept. of Agronomy) and with state and federal soil scientists was implemented to develop statistically-based procedures for collecting information on soils and their properties. The project integrates geographic information systems (GIS) and global positioning systems (GPS) to randomly select and to locate sample points. A multi-phase design based on Markov chain sampling procedures was implemented to balance statistical and operational constraints. Research involved developing estimation procedures to summarize distributional properties of soil map units, with collaborators Jay Breidt (Colorado State University) and Pamela Arroway (North Carolina State University).

Estimating usual dietary intake distributions

CSSM faculty collaborated with Alicia Carriquiry (Statistics) and Helen Jensen (Dept. of Economics and Center for Agricultural and Rural Development [CARD]) to develop methods for estimating usual dietary intake distributions from survey data consisting of daily dietary intakes from individuals. These data contain nuisance effects, are non-normal, and contain substantial measurement error. A semi-parametric transformation coupled with a measurement error model framework are used to estimate usual intake distributions. Research involved extending the method developed for nutrient intakes to accommodate food intake data and working with the National Center for Health Statistics to apply these methods to blood measurements.

Nitrogen runoff prediction

Jean Opsomer collaborated with Bruce Babcock (Dept. of Economics and CARD) to develop a mixed parametric/nonparametric statistical model for nitrogen runoff prediction, in order to evaluate the environmental impacts of agricultural policy scenarios.