Nonsampling Error Analysis
Our researchers and statisticians have developed innovative and effective strategies to minimize the occurrence of avoidable errors and minimize the effects of unavoidable errors in clinical trials, experiments, and surveys. Our analytical techniques address the bias that results from measurement or data collection errors, nonresponse and other missing data, the influence of interviewers or clinicians, and respondent recall errors.
Focus Areas
Nonresponse Reduction and Compensation
- Interviewer training in nonresponse avoidance methods
- Refined and tested methods for respondent informed consent and assurances of confidentiality
- Use of incentives
- Private or anonymous data collection modes
- Development of new weight adjustment methods
- Methods for imputation that appropriately reflect imputation uncertainty and model error
Evaluation and Reduction of Survey Error
- Survey designs to minimize occurrence of measurement errors
- Experimental designs to allow measurement error components to be quantified
- Advanced interviewer training and supervision methodologies
- Cognitive laboratory pretesting methods
- Designs and statistical methods for comparing alternative modes of data collection
- Statistical models for nonsampling error structures
- Analysis of unit and item nonresponse bias after weighting and imputation, respectively
Measurement Error Modeling
- Re-interview survey design for estimating reliability and response bias
- Models for studying measurement bias under complex sampling schemes
- Mutlilevel models for describing effects of interviewers, coders, and other survey personnel on survey response
- Probability models for assessing validity of self-reported drug use
- Models for describing effects of measurement and nonresponse errors on categorical data
Reduction of Data Processing Error
- Quality control and quality assurance methods attached to all data processing activities
- Analysis of key process statistics to reduce the risk of keying, editing, coding and other processing errors
- Quality circles and continuous quality improvement methodology
- Recruitment, training and retention of highly capable data processing personnel
Methodologies/Techniques
- Statistical methods for mode comparison studies
- Latent class analysis
- Structural equation modeling
- Item response theory
- Test-retest reliability analysis
- Estimation of internal and external validity
- "Gold standard" studies for bias evaluations
- Quality assurance and quality control methodologies
- Total survey error analysis