Studying the earthquake processes involves the exploration of the available data to extract all possible information required for the deeper understanding of the underlying physical mechanisms. When a theoretical concept, in the form of a physical and/or statistical model, is developed and applied, data of the best possible quality (in terms of precision, sample size etc) is essential for the adequate estimation of the required parameters. In the present work, issues relevant to data sets extracted from earthquake catalogues and their effect on seismicity studies are discussed, based on simulations of seismicity data. For example, the realizations of data based scenarios presented here for the investigation of earthquake predictability give evidence of a potential use of foreshocks in earthquake forecasting, if more data becomes available. Assuming that an ideal improvement of the seismic networks can be achieved, an increase of network sensitivity by 1.5 magnitude unit would enable us to correctly detect 4 times more foreshock sequences. The corresponding number of false detections would also increase but remain relatively lower. However, real data sets can be far from perfect even if numerous earthquakes are recorded, and results might be misleading if e.g. the magnitude of completeness is underestimated due to errors in the reported magnitudes. The analyses of stochastic magnitude catalogues which include data-based errors reveal and quantify the significant impact the quality and size of a data set might have on the estimated catalogue properties, and as a result on seismic hazard assessment.