The aim of the MICROSCOPE space mission is to test the weak equivalence principle (WEP) with a precision of 10⁻¹⁵, 100 times better than on-ground experiments. The WEP violation signal that we look for in flight has a particular frequency signature and must be extracted from the measurement of accelerations dominated by a colored noise integrated over several orbits. The measured data include various modeled or possibly unknown perturbation signals and may be frequently interrupted during the long integration period. Therefore the cautious analysis of the time series that will be obtained is of utmost importance for the validation of the mission. We present a linear regression procedure that we have developed to tackle the problem of missing data in the presence of correlated residuals. While the uncertainty of standard least square methods is increased by more than an order of magnitude in the presence of gapped data, we show that our method allows us to maintain the targeted level of precision of the experiment. We finally test the implemented tools on simulations of the in-orbit instrument with an end-to-end simulator of the mission.