The performance of a photovoltaic (PV) system depends on several factors, such as the solar radiation availability and its spectral distribution, the PV module temperature, soiling, cable losses, PV power degradation over time and so forth. An important factor that also affects the PV array power is the mismatch loss due to the differences between single modules, since is inherent to the manufacturing process certain variability in the I-V curve parameters. The manufacturing technology of PV modules has improved considerably, resulting in higher efficiencies and better quality control process, which enabled a lower maximum power tolerance range of PV modules available in the market. The actual shape of the statistical distribution of the main electrical parameters is necessary to evaluate the mismatch losses using simulation software, and also to verify if a new selection of PV modules besides the one performed by the manufacturer is relevant. In order to analyze these topics, a statistical study was carried out based on data obtained from I-V curve measurements of 105 multicrystalline PV modules with the same nominal characteristics. The measurements were performed in a pulsed solar simulator in standard test conditions. The descriptive statistics were obtained for each main electrical parameter and the best probability density function that describes the parameters dispersion was determined. The results show that the maximum power, the maximum power voltage and the open circuit voltage are preferably represented by a Burr probability density function, however a normal distribution is adequate as well. The short circuit current, the maximum power current and the fill factor are actually described by a two parameter Weibull distribution. In order to analyze the effects of the mismatch losses in arrays, several I-V curves of strings with 10 PV modules randomly selected from the sample were synthesized and compared to strings of modules sorted by the maximum power current value. The advantage of performing a new selection of PV modules with better current match was not relevant in comparison to random strings. The selection performed at the factory for a PV module with the same nominal power is sufficient to prevent considerably mismatch losses considering that the PV modules were sorted using standard procedures. (C) 2016 Elsevier Ltd. All rights reserved.