Context. One possible scenario to explain the emission from the stellar binary system LS I +61• 303 is that the observed flux is emitted by precessing jets powered by accretion. Accretion models predict two ejections along the eccentric orbit of LS I +61• 303: one major ejection at periastron and a second, lower ejection towards apastron. Our GeV gamma-ray observations show two peaks along the orbit (orbital period P 1 ) but reveal that at apastron the emission is also affected by a second periodicity, P 2 . Strong radio outbursts also occur at apastron, which are affected by both periodicities (i.e. P 1 and P 2 ), and radio observations show that P 2 is the precession of the radio jet. Consistently, a long-term modulation, equal to the beating of P 1 and P 2 , affects both radio and gamma-ray emission at apastron but it does not affect gamma-ray emission at periastron. Aims. If there are two ejections, why does the one at periastron not produce a radio outburst there? Is the lack of a periastron radio outburst somehow related to the lack of P 2 from the periastron gamma-ray emission? Methods. We develop a physical model in which relativistic electrons are ejected twice along the orbit. The ejecta form a conical jet that is precessing with P 2 . The jet radiates in the radio band by the synchrotron process and the jet radiates in the GeV energy band by the external inverse Compton and synchrotron self-Compton processes. We compare the output fluxes of our physical model with two available large archives: Owens Valley Radio Observatory (OVRO) radio and Fermi Large Area Telescope (LAT) GeV observations, the two databases overlapping for five years. Results. The larger ejection around periastron passage results in a slower jet, and severe inverse Compton losses result in the jet also being short. While large gamma-ray emission is produced, there is only negligible radio emission. Our results are that the periastron jet has a length of 3.0 × 10 6 r s and a velocity β ∼ 0.006, whereas the jet at apastron has a length of 6.3 × 10 7 r s and β ∼ 0.5. Conclusions. In the accretion scenario the observed periodicities can be explained if the observed flux is the intrinsic flux, which is a function of P 1 , times the Doppler factor, a function of β cos( f (P 2 )). At periastron, the Doppler factor is scarcely influenced by P 2 because of the low β. At apastron the larger β gives rise to a significant Doppler factor with noticeable variations induced by jet precession.