A Note on the Effect of Ignoring Small Measurement Errors in Precision Instrument Calibration.
Abstract
The authors' focus is the simple linear regression model with measurement errors in both variables. It is often stated that if the measurement error in x is small, then we can ignore this error and fit the model to data using ordinary least squares. There is some ambiguity in the statistical literature concerning the exact meaning of a small error. For example Draper and Smith (1981) state that if the measurement error variance in x is small relative to the variability of the true x's, then errors in the x's can be effectively ignored, see Montgomery & Peck (1983) for a similar statement. Scheffe (1983) and Mandel (1984) argue for a second criterion, which may be informally summarized that the error in x should be small relative to (the standard deviation of the observed Y about the line)/(slope of the line). We argue that for calibration experiments both criteria are useful and important, the former for estimation of x given Y and the latter for confidence intervals for x given Y. Keywords: Confidence intervals. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1985
- Accession Number
- ADA159104
Entities
People
- C. H. Spiegelman
- Raymond J. Carroll
Organizations
- University of North Carolina at Chapel Hill