Propagation of variance uncertainty calculation for an autopsy tissue analysis

An assay value is almost useless unless it is accompanied by a measure of the uncertainty associated with that value. The uncertainty value should reflect all the major sources of variation and bias affecting the assay and should provide a specified level of confidence (for example, 95%). An approach to uncertainty calculation that includes the uncertainty due to instrument calibration, values of the standards, and intermediate measurements as well as counting statistics is presented and applied to the analysis of an autopsy tissue. This approach, usually called propagation of variance, attempts to clearly distinguish between errors that have systematic (bias) effects and those that have random effects on the assays. The effects of these different types of errors are then propagated to the assay using formal statistical techniques. The result is an uncertainty on the assay that has a defensible level of confidence and which can be traced to individual major contributors. However, since only measurement steps are readily quantified and since all models are approximations, it is emphasized that without empirical verification, a propagation of uncertainty model may be just a fancy model with no connection to reality.