An extensive work on the luminosity calibration of the HR diagram has been performed using the LM method based on the Maximum Likelihood estimation (Luri et al. 1996). The method uses all available information: Hipparcos parallax and proper motion data as well as radial velocities and some measured spectroscopic or photometric parameters related to luminosity. On the other hand, the method takes into account the censorship of the sample and the errors of the data. Finally, it is able to identify and separate in a given sample, groups with different luminosity, kinematical or spatial characteristics providing not only the corresponding luminosity calibration relationship, but also improved individual absolute magnitude estimates.

In order to asses how reliable the luminosity calibrations based on the MK spectral classification are, the LM method has been applied to a sample of B to K stars with MK classification selected from the Hipparcos survey. All known binaries, variables and spectroscopically peculiar stars were removed from the sample. The final sample contains 22054 stars. Our results show that the relation between absolute magnitude and luminosity class has a large intrinsic dispersion. As a consequence, the assignation to a star of a mean absolute magnitude given by a MK classification calibration is a rough procedure. It is preferable to use the individual absolute magnitudes estimated using all the available information for each star, like the estimates provided by the LM method (Gómez et al. 1997).