Bio-Rad Microplate Manager Software - Manual (Page 108)

Microplate Manager User Guide
98
If there is a significant difference between standard curves that should be essentially
identical, the user should investigate the source of this difference.
Note that these statistics depend on the type of curve fit chosen. Changing the
regression method and axis type will change the analysis of variance.
This information will also be printed at the bottom of the Standard Curve Report.
An exhaustive discussion on the analysis of variance is beyond the scope of this
manual. The interested user is referred to Biometry by Robert R. Sokal and F. James
Rohlf, W.H. Freeman and Company, New York, 1981.
6.8 Standard Curve Fitting
The Microplate Manager program provides powerful standard curve fitting and data
reduction routines. You can generate the standard curves by five different methods:
Linear, quadratic and cubic polynomial regression, cubic spline, and four-parameter
logistic curve fitting. For the first four methods, the axes may also be in four different
combinations of linear and logarithmic (i.e., linear-linear, log-linear, linear-log and
log-log); the logistic curve fit is constrained to linear-linear axes. When used
properly, these can greatly enhance and simplify quantitative microplate assay
evaluations.
Overview of Regression Methods
Polynomial regression and logistic curve fits are examples of global construction of a
calibration function, while the cubic spline is an example of a local approach. The
global approach involves using the entire set of standards for calibrating, while the
local approach employs a varying subset of the standards. The global approach
provides better noise immunity and allows for minor extrapolation when a calibration
function is used to analyze a physical process whose mathematical form is well
established in the full range or at least the range of interest.
Cubic regression as a global function is useful for interpolating nonsymmetrical
functional dependencies with an inflection point (such as the plot of migration
distance versus the logarithm of the molecular weight). Higher orders of polynomial
regressions are rarely used for calibration if a good fit is not found with linear,