in development
This will be a short tutorial, but it can be skipped if you are confident the residuals from your calibration regression model will be homoscedastic. This may be the case if the mean variance of the response does not increase with increasing calibrator level, or if some variance-stabilising function of the response is conventionally and consistently used (and well-understood) for your test type.
For the rest of us whose test responses are inherently or possibly heteroscedastic, some preliminary work can help with our later regression model choices.
Homoscedasticity (homoskedasticity; homogeneity of variance):
An assumption of linear regression
“At every value of \(x\), the variance of \(y\) is the same!”
(Kruschke 2011, 420)
The opposite of heteroscedasticity
“In practice of course, … it will almost certainly be related to the expected response, i.e. the error will not be constant at all points on the curve. This non-constancy of error is referred to as heteroscedasticity and has important implications when fitting calibration functions to responses from standard series.”
(Law et al. 2005, 173)
Kruschke, John K. 2011. Doing Bayesian Data Analysis: A Tutorial with R and BUGS. Academic Press, Elsevier, USA. http://www.indiana.edu/~kruschke/DoingBayesianDataAnalysis/.
Law, Brian, Clive G Copley, Robert A Biddlecombe, Michael J Warwick, Michael D Malone, and William J Jenner. 2005. Immunoassay: A Practical Guide. Edited by Brian Law. e-book. Taylor & Francis.