The accuracy of the structural parameters (i.e. type of neighbour, number of neighbours of a specific type, distance between absorber and neighbour, and disorder) that are obtained from the analysis of EXAFS spectra is in theory only depending on the errors in the data. However, the experimental set-up, the standard used to build the model EXAFS, and data analysis method may introduce systematic errors. This paper discusses several methods to decrease the errors associated with the method of data analysis. With MoS2 as an example the use of phase corrected Fourier transforms to determine the type of neighbour is demonstrated. The conventionally applied k(3)-weighting of the data was compared to k(1)-weighting and weighting by the inverse of the statistical error in the refinement of the structural parameters of PtO2. Weighting the data with the inverse of the statistical error gives results in agreement with XRD structure determinations and the best goodness of fit values (epsilon(v)(2)).