Bio-Rad Microplate Manager Software - Manual (Page 109)

Chapter 6. Data Formatting and Reports
99
quadratic, or cubic polynomial regression, because of the oscillatory nature of the
resultant fit. In these cases, a local fit (cubic spline) would be preferable.
The concentration of unknowns is determined by substituting the mean absorbance
values of the unknowns in the regression equation and solving for the concentration.
Microplate Manager performs data reduction without artificial limitations or
constraints. This provides added flexibility, but leaves the final interpretation of the
results to the user.
In general, you should start with the simplest curve fitting routine (i.e., linear) and
examine all four combinations of axes for an acceptable curve fit. If a satisfactory fit
is not obtained, then progressively increase the sophistication of the curve fitting until
a good fit is obtained.
Ways to Increase Reliability
There are several approaches that you can take to increase the reliability of the
results:
•
Run an adequate number of standard dilutions to form a broad dose response
range. Twelve standards of 1:2 dilutions provide a 3.5 log range of standards,
which is usually adequate for most applications.
•
Run at least three replicate wells for each dilution of standards and samples.
•
Run an adequate number of sample dilutions to ensure that more than one
dilution lies in the middle of the curve. Running several dilutions will point out
the limitations of the curve-fitting methodologies.
Goodness of Fit
The correlation coefficient (r) gives an indication of how well the polynomial
regression (linear, quadratic and cubic) model fits the data. In effect, the correlation
coefficient is the square root of the proportion of explained variation to total
variation of the regression.
r = (explained variation / total variation)
1/2
The amount of explained variation increases as the goodness of fit increases. In a
perfect fit, all of the variation is explained, the explained variation equals the total