3.3.4. Data analysis
The use of the ImmunoAssay Calculations spreadsheet freely available from Bachem (http://www.bachem.com/service-support/immunoassay-calculator) is highly recommended. This program is designed to run a four parameter non-linear regression analysis of RIA data.
In this spreadsheet the mean of the non-specific binding counts (NSB) is first subtracted from the means of all other samples. Next, each of the standard curve means (B-NSB) is divided by the maximum count (B-NSB/B0-NSB) and these values are then fitted to a nonlinear regression curve given as:
where a represents the maximum binding value (set to 1), b the slope, c the inflection point of the curve (lc50) and d
the minimum value. Through a graphic conversion of the minimization procedure, the spreadsheet makes this fitting a user friendly task. Finally, sample values are entered and converted to amounts of hormone in each sample.
Such non-linear regression is far superior over a linear log/logit regression analysis for calculating hormone concentrations in samples close to the detection limit of the RIA.
An important quality check for all radioimmunoassays is the calculation of intra- and inter-assay variation. Whereas intra-assay variation usually reflects the level of pipetting errors, inter-assay variation is of relevance when large sample series are processed or when results are to be compared to other studies. This is done by including in each assay a sample prepared from an aliquot of a large haemolymph-pool sample. As this sample will be included in any of the other assays done on the same species, it allows for quality control of the RIA and to assess inter-assay variation.