Fig 1.
Example for summarizing the ELISA dilution absorbance data using three methods.
The points are absorbance values (red numbers) at different dilutions from a sample in the dataset. The sigmoidal curve is a fitted 4-parameter model with the point of maximum growth shown as green *. The endpoint titer value is represented by the blue point above the blue dashed horizontal line. The absorbance summation is the sum of all the red numbers.
Fig 2.
Examples of successful and unsuccessful sigmoidal curve fitting.
A. Successful sigmoidal curve fit where the data cover the estimated point of maximum growth (PMG). B. Successful sigmoidal curve fit even though data are not available for the estimated PMG. C. Unsuccessful sigmoidal curve fit due to the lack of informative data for fitting the curve. Line, fitted sigmoidal curve.
Fig 3.
Linear relationship between the absorbance summation (AS) and estimated point of maximum growth (PMG).
For observations with an estimated PMG within the observable concentrations, the AS and the PMG have a clear negative correlation. The straight line is the regression line.
Fig 4.
Relationship between the three proxies and simulated PMG.
Sigmoidal curves are simulated with different point of maximum growth (PMG). Each estimated proxy is plotted against the simulatied PMG. The estimated PMG works well for values where there is more data surrounding the PMG, but variance increases when the available data are not near the PMG. The endpoint titer method results in discrete values and has a weaker relationship with the actual PMG. The absorbance summation has a strong relationship with the true PMG.
Fig 5.
Statistical power comparison among the three methods.
Statistical power is plotted against the PMG differences between the two comparing groups. Sigmoidal curves are simulated with different PMG, which range from within (-16.0959 and -15.0959) to outside of the measurements (-13.0959 and -11.5959), as shown in the panel columns. The error variance is either constant or changing along the curve with a linear or quadratic relationship with the values on the curve, which are shown in different panel rows. Statistical power is calculated based on the two group t test. Var, variance.
Fig 6.
Statistical power comparison between AS and AUC.
Statistical power is plotted against the PMG differences between the two comparing groups. Sigmoidal curves are simulated with different PMG (-16.0959, -.0959 and -24.0959 for the three panel columns). The variance for each measurement of the simulated curve is either constant or changing along the curve with a linear or quadratic relationship with the measurement values on the curve in the three panel rows. Statistical power is calculated based on the two group t test. AUC is estimated based on quadratic curve fitting and integration. Var, variance.