Random measurement error in an outcome variable tends to increase the standard error of the estimates (SEE), with widening of the confidence intervals (CI). There is a loss in precision which can reduce the apparent statistical significance of the variable.
Ways to reduce the error:
(1) increase the sample size
(2) increase the number of repeated measurements taken per subject
If the reliability coefficient for the measurement (a decimal fraction) is known:
number of subjects to enroll in the sample when there is a measurement error =
= (number of subjects required if there was no measurement error) / (reliability coefficient)
This calculation can be used to determine how much smaller a sample size could be if a more reliable measurement method was used.
The number of repeated measurements required to achieve a certain level of precision can be calculated using the Spearman-Brown formula.
