ABSTRACT The full energy peak efficiencies and covariances of a high purity germanium (HPGe) detector studied by the regression method and Monte Carlo method were compared. The impact of the obtained results on the neutron activation cross-sections measured relative to monitor cross-sections was studied. In the regression method, the efficiencies measured for a set of calibration point sources were analysed by the least-squares analysis. In the Monte Carlo method, the efficiencies for the calibration point sources were calculated by MCNP. The covariances of the efficiencies determined by the regression method were calculated analytically. Perturbation analysis was performed to estimate the covariances of the efficiencies calculated by the Monte Carlo method. Positive correlations higher than 0.8 were found in the uncertainties of the MCNP data for point-like sources. In the case of the regression method, the correlation coefficient contains both positive and negative terms. The efficiencies and their covariances for finite sample geometry were also estimated by both methods to take into account sample effects such as geometrical effect, and gamma-ray self-absorption, and found considerable differences in the cross-sections and their uncertainties for reaction products quantified with low energy gammas. The efficiencies and covariances were clearly affected by the properties of the sample.