4. Conclusion

We have formalized a bioassay experiment in terms of statistical modeling. The appropriate models are non-linear regression models. The methods used are based on statistical test theory. Several methods for calculating confidence intervals for model parameters or calibration intervals have been presented. Unfortunately, the statistical properties of these methods are only known under the assumption of a large number of observations, and nothing can predict in advance which method will give the best results for a particular experiment where the number of observations is limited. The experimenter therefore has several methods at his disposal, from which he can draw to support his conclusions. This dossier only gives an overview in the particular case where observations are assumed to have a Gaussian distribution and all have the same variance. If the variance of the observations depends on...