In this thesis, we tackle several open problems in the study of large-scale structure through the clustering of galaxies. To this end, we analyse data from some of the latest surveys, and we also develop new statistical techniques needed for this analysis in specific cases. In the first part, we focus on small and intermediate scales, where the relation between galaxy properties and their clustering (known as segregation) is important. The work in this part is driven by the exploitation of data from the ongoing Advanced Large Homogeneous Area Medium-Band Redshift Astronomical (ALHAMBRA) survey, which is perfectly suited to study the evolution of segregation trough cosmic time. We developed a method for the recovery of the real-space clustering from photometric surveys with the characteristics of ALHAMBRA. This method is based on the use of the projected correlation function, and is adapted to data with typical photometric redshift errors Dz < 0.015 (1+z). We tested the method using N-body simulations, and then applied it to the calculation of the correlation function for several samples drawn from the ALHAMBRA survey. We divided our sample in three redshift bins, and selected several galaxy samples in each of them based on B-band luminosity. In the range of scales studied, the correlation function for all samples was well fitted by a power law. We observed as well the effects of evolution, and of luminosity segregation. We also reviewed the basic tools available in the framework of the statistics of marked point processes to study galaxy segregation. We illustrated their application using a galaxy sample drawn from the 2dFGRS, characterised by a spectral classification parameter. We introduced the mark connection function, showing that it gives valuable information when analysing different galaxy populations defined by some set of galaxy characteristics. In the second part, we focus on the study of a large scale feature of the galaxy distribution, the baryon acoustic oscillations (BAO). We measured the two-point correlation function for several samples drawn from the largest surveys to date, 2dFGRS and SDSS. We obtained a peak corresponding to BAO at the expected scale in all cases, which shows the reliability of the detection of this feature. Finally, we developed a new method for the analysis of the BAO phenomenon. This method makes use of the possibilities of wavelets methods to look for the actual structures in configuration space which are responsible for the BAO. It is also based on the use of two complementary mass tracers, and we illustrated it using a catalogue formed by `Main' and Luminous Red Galaxy samples from SDSS. In this way, we showed how we were able not only to detect BAO in the samples, but also to localise regions giving lower or higher BAO signal. This kind of information is completely lost when using the traditional two-point statistics methods.