This paper provides a comparative analysis of statistical algorithms for the retrieval of the wet tropospheric correction from microwave radiometers in the context of altimetry missions. The algorithms are based on the algorithms used in Envisat and Jason-1 missions. The objective of this comparison is two-folds:1) To find which regression method is better suited for the retrieval between the neural network algorithm and the log linear regression.2) To tackle the problem of variable selection, i.e. to find the best set of variables to include as inputs in order to reduce the retrieval error. In particular, we want to determine whether the lack of a radiometer third channel at 18GHz can be compensated by the altimeter backscattering coefficient.Several configurations of algorithms, including those used in the operational processing of altimetry missions such as JASON-1 or ENVISAT, are built and compared on the same learning and test databases to determine which retrieval strategy is more appropriate. The database is composed of atmospheric and oceanographic conditions taken from the European Centre for Medium-Range Weather Forecasts (ECMWF) analyses and the brightness temperatures are simulated using a radiative transfer model. The importance of each input for the different algorithms is analyzed and the performances of the different algorithms are assessed in terms of error (bias and standard deviation) but also in terms of geographical distribution of the errors and correlation with other environmental variables. The results are then validated on Jason-2 radiometer measurements.Our results show that, in terms of variable selection, better results were obtained when the brightness temperature at 18 GHz was used instead of the backscattering coefficient. Moreover, better estimations of the wet tropospheric path delay were obtained with neural networks.