The assertion in the title requires only that data be inverse-variance weighted in the calibration fit, in which case the standard errors (SEs) at the heart of these limits are almost exact for any algebraically linear response function. Furthermore, for quantities in the response domain, these SEs require no actual calibration data; and only the calibration slope is needed to translate these into the concentration domain. This means that calibration models can be designed to yield specified values of these limits — a simple reality that seems underappreciated by analytical workers. Additionally, simple algorithms permit numerical evaluation of the SEs for any response function, facilitating calibration with the nonlinear response functions that arise in many analytical methods. The catch, of course, is the data variance function (DVF), which can be known exactly only in the context of Monte Carlo simulations. This work will emphasize the determination of DVFs in calibration, with a focus on how good is "good enough."

Calibration-based detection and confidence limits are (almost) exact when the data variance function is known

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.