Behavioral Risk Factor Survey Relative Standard Error

Survey results are estimates of population values and always contain some error because they are based on samples. Confidence intervals are one tool for assessing the reliability, or precision, of survey estimates. Another tool for assessing reliability is the relative standard error (RSE) of an estimate. Estimates with large RSEs are considered less reliable than estimates with small RSEs.

How large is "large" when looking at RSE? There is no absolute cutoff point. The Office of Health Informatics follows guidelines used by the National Center for Health Statistics (PDF, 289 KB, exit DHS) and recommends that estimates with RSEs above 30 percent should be considered unreliable.

How is RSE calculated? Relative standard error is calculated by dividing the standard error of the estimate by the estimate itself, then multiplying that result by 100. Relative standard error is expressed as a percent of the estimate. For example, if the estimate of cigarette smokers is 20 percent and the standard error of the estimate is 3 percent, the RSE of the estimate = (3/20) * 100, or 15 percent.

How should RSE be applied to the estimates produced in this module? RSE is particularly helpful where the confidence interval is quite large; for example, 8%-9% or larger. In such a case, the reliability of the estimate would be suspect in the absence of additional information; however, if RSE does not exceed 30%, the estimate may still be considered reliable.