The construction of the proper motion system of the compiled catalogue was carried out in the usual way. Systematic differences needed for this purpose have been derived through pair-wise comparison of individual catalogues. Preliminary, magnitude dependent errors of proper motions were redetermined for each initial catalogue. Usual statistical methods were applied to this purpose. In such case systematic differences do not depend on brightness of stars and become less than 3 mas/yr.
The investigation of systematic differences has been performed by the analytic method developed at Heidelberg (Schwan 1988). The systematic relations between catalogues were represented by a series development using orthogonal functions. The orthogonal functions adopted in this work are the products of Legendre polynomials, Hermite polynomials and Fourier terms. Using the transformed declination this type of functions is the most suited for modelling the systematic differences between catalogues under consideration. Only the significant terms of the series development were actually used. The F-test with the level of five percent was used for this purpose. The analysis shows that systematic differences between catalogues depend on right ascension and declination and do not depend on magnitude. The significant terms and their rms errors were applied for deriving a common system of proper motions. In order to do it we were guided by the idea that the weighted mean system is the most reliable one. The weights of catalogues have been assigned according to their consistency with the mean system. The proper motions of initial catalogues have been reduced to the adopted system by applying quite small corrections, with values less than 4 mas/yr.
After the proper motion differences had been corrected for the systematic part, we investigated the random errors in each initial catalogue. It was found that the mean errors are different for individual catalogues with values between 4 and 12 mas/yr. The weighted mean proper motions were calculated for common stars using the weights assigned to each catalogue as the inverse square of the random errors.