Errors-in-variables model Created 2026-10-14 Updated 2026-11-05
An Errors-in-variables model distinguishes a latent true predictor from its noisy measurement and models both the measurement and the response. For example, and , with a population distribution for , propagate predictor uncertainty into inference on . Treating as exact can create attenuation bias from classical measurement error. Identifiability requires appropriate information about the measurement process or latent predictor distribution.
Measurement error Created 2026-10-03 Updated 2026-11-05
A measurement error is the difference between a recorded measurement and the quantity intended to be measured. The additive model separates that error from genuine variation in . A shared offset is a systematic error; fluctuations between repeated measurements are random measurement error. Repeating a measurement can reduce independent Random measurement error, but does not remove a shared systematic error.
measurement error in a predictor can bias a regression slope even when it has mean zero. Suppose , where is uncorrelated with and , and is uncorrelated with . A regression of on the recorded has population slope
Thus equal predictor and error variances halve the slope: this is attenuation bias from classical measurement error. In contrast, independent mean-zero error added only to increases residual variance without changing that population slope. An errors-in-variables model describes the measurement mechanism explicitly rather than treating noisy predictors as exact.
The baseline assay is a noisy measurement just like the follow-up assays. On the log scale, a normal distribution for its error is a reasonable approximation if variability is approximately multiplicative on the original titre scale. The latent represents the true baseline log-titre, with
Using a common variance is a simplifying assumption justified if baseline and follow-up measurements have comparable assay variability; it should not be treated as a fact guaranteed by the study design. Treating as exact ignores measurement error in a predictor and can lead to attenuation bias from classical measurement error and understated uncertainty.
This is attenuation bias from classical measurement error, also called regression dilution. The latent predictor is , but and are noisy proxies. The noisier proxy produces a slope closer to zero.
For a regression with an intercept, the population slope is
Here all variables are centred and the noises are independent. Therefore
so regressing on has slope . For
we have
Thus
With the stated values, the slopes are
matching the simulation.
Under this independent additive measurement-error model, the magnitude of the -on- slope is generally smaller than the magnitude of the -on- slope whenever . Increasing does not change the population slope, because response noise contributes neither to nor to . It increases residual variance and the sampling variability of the estimate; with one million observations, doubling it should leave the displayed slope close to while increasing its standard error.