Concepts

Comparison precision

Selecting the optimal comparison precision is a challenge. It depends on the sensitivity and properties of your assay method and may vary from assay protocol to assay protocol.

Comparison precision defines the number of significant digits that are taken into account when comparing assay results. Due to differences in the way computer calculations are carried out with IEEE mathematics and due to differences in the algorithms, it is normal and expected that calculation results may vary slightly. While this is not critical, comparisons carried out with full precision will reveal these slight differences.

Differences are always expected when you perform nonlinear regressions (e.g. 4- and 5-parameter fits) since the precision of these methods has been improved within PLA 3.0.

To deal with the differences all comparison variables have been divided into classes:

• Regression parameters
• Test results
• Potency results
• Standard errors

These classes have been shown during the development to potentially require different precision settings. You will need to assess how precise is precise enough for your system. By default, all classes except for the standard error compare with 5 significant digits and the standard error class due to the sensitivity of non-linear curve fits compare with 3 digits.

Comparison result

The system will not state that the comparison is failed or passed. It will just state, whether with the selected precision settings deviations have been detected or not. It is a scientific decision to accept these values or not.