The article describes the methods for evaluation of the basic parameters using the least squares (zero offset, slope of the output characteristics (scale factor), and so on) of the micromechanical transducers of linear acceleration (PLD). It also contains studies of the static and random errors of the micromechanical transducers of linear acceleration (accelerometers) and random errors of the measurement parameters using the least squares metho, and Alan variations. Due to minimization of the experimental and theoretical values of the measured parameters and the least squares method it became possible to significantly reduce the measurement error. The accuracy of the measured values of the accelerometer using Alan variations was significantly improved due to an increase of the number of partitions of points in the group. So, in the split points in group 51 the accuracy was equal to 10 % and in group 201 to 5 %. The authors considered two ways for estimation of the parameters of a linear acceleration sensor: a static one, using centrifuges, and a dynamic one with testing on a shaker. As is known, a linear acceleration can occur due to the rectilinear, curvilinear and rotational movements. Since in this case the subject of interest is the linear acceleration, the play will be performed using the rotational motion of the centrifuge. The rotational motion and the linear motion depend on the distance of a product to the center of rotation of the table and the speed. The measurement result will depend on the values of the device parameters and characteristics of the test modes of a centrifuge. Because the effects of vibration of a product may cause mechanical defects or deterioration of the values of the parameters, of great importance is vibration testing on a vibration table. The main requirement to the test bench is a reproduction of vibration in a predetermined frequency range and mode for a set time and with a given accuracy.