The thesis deals with the identification of the optimal adjustment models of Scitnrex CG-3M gravimeter’s measurements with respect to the most significant instrumental error influences. On the basis of available literature and experience with two CG-3M gravimeters, the hysteresis after transport, transport drift and errors in the calibration function are identified as the most significant instrumental error influences in Scintrex CG-3M measurements. The analysis of the empirical data (comprising four sets of different purposes) in the phase of pre-processing showed that the two CG-3M gravimeters are affected with the significant hysteresis effect. An iterative algorithm has been developed, which models hysteresis using exponential function and facilitate its elimination from observation series of sufficient duration. In addition, the analysis showed that the gravimeters exhibit different, but rather regular, drift behaviour during different types of the survey. Furthermore, an uncertainty of determined calibration coefficients or its time change can significantly affect the relative gravity measurements. The problems of datum definition and measures of quality for gravity networks are analysed and resolved. Different adjustment models with respect to the treatment of the most significant error influences in the functional and stochastic part of the adjustment model have been tested on the empirical data. The investigation showed that the corrections of the linear calibration coefficients and coefficients of the second order drift polynomial should be, in general, included in the functional model as parameters. In addition, the hysteresis can significantly affect the precision of the Scintrex measurements, which cannot be remedied by a stochastic model.