A complete statistical model taking into account the nature of the magnetic sensor noise is proposed for parameter extraction in magnetorelaxometry experiments on magnetic nanoparticles (MNPs). Apart from the noise, the important factors are the type of the magnetic sensor used as well as the regression model depending on the relaxation mechanism. Whatever the relaxation mechanism is, the models are non-linear. Consequently, the relation between the signal to noise ratio of the measurement and the uncertainty of the extracted parameters is not straightforward. From the model, we derive a figure of merit that enables an a priori performance extrapolation from the characteristics of the magnetometer.