|
|||||
|
|
Assessing the impact of measurement error in modeling change in the absence of auxiliary data |
|
| Authors: | N. David Yanez Ibrahim Aljasser Mose Andre Chengcheng Hu Michal Juraska Thomas Lumley |
| Affiliation: | 1. Department of Biostatistics, University of Washington, Seattle, WA, USA;2. Department of Quantitative Analysis, King Saud University, Riyadh, Saudi Arabia;3. Surefield, Seattle, WA, USA;4. Epidemiology and Biostatistics Division, University of Arizona, Tucson, AZ, USA;5. Biostatistics and Biomathematics, Fred Hutchinson Cancer Research Center, Seattle, WA, USA;6. Department of Statistics, University of Auckland, Auckland, New Zealand |
| Abstract: | Measurement error is well known to cause bias in estimated regression coefficients and a loss of power for detecting associations. Methods commonly used to correct for bias often require auxiliary data. We develop a solution for investigating associations between the change in an imprecisely measured outcome and precisely measured predictors, adjusting for the baseline value of the outcome when auxiliary data are not available. We require the specification of ranges for the reliability or the measurement error variance. The solution allows one to investigate the associations for change and to assess the impact of the measurement error. |
| Keywords: | Errors in variables linear regression measurement error variance measurement reliability method of moments sensitivity analysis |