Nonstandard deviations.

Abstract

An estimate of uncertainty without a statement as to whether the estimate is a standard deviation or related quantity or an indication of the associated confidence level is meaningless. Nevertheless, the necessary qualification is often omitted. While such omissions are probably most often inadvertent, they may occur because the conventional single standard deviation of a normal distribution is inappropriate. Some unusual uncertainty intervals are therefore suggested. These include noninteger multiples of the standard deviation such as 1.5 standard deviations, asymmetric error bounds, and identification of the possibility of long tails or truncated bounds to the probability density function.