[Complementary methodological steps in the analysis of rhythms: search of periods, modelling. Examples of plasma melatonin and temperature curves].

The choice of the period is a problematic in the field of the analysis of rhythms which is not yet sufficiently formalized in the sense that it was often postulated that the period is known and more precisely in the case of the circadian rhythms. We propose some methods which have a great reliability in the determination of the period and consequently can allow to characterize the parameters of the rhythmic phenomena (amplitude, phase, rhythmic level) with more precision. Traditional "cosine" modeling is always an unquestionable interest in the sense that it is a good reflect of the behavior of a biological periodic phenomenon in the course of time. However its employment requires to check complementary assumptions of validity in front of a statistical expertise. We will show the main ones. On the other hand it seems that this model requires a certain adaptation as it shows for example in the variabilities of the phase that one can notice in the behaviour of the biological rhythmic phenomena.