Current estimations of cosmological parameters depend often on assumptions about the cosmological model itself. The estimates of $H_{0}$ or $\Omega_{m}$ from CMB surveys, for instance, are valid only assuming a particular model, typically the standard $\Lambda$CDM. Similar model-dependent results are obtained also from analyses of large-scale structure. In some case it is however possible to combine observations in such a way to get estimates of physical quantities that are valid regardless (to some extent) of the underlying model. Here I will discuss how one can determine the cosmological expansion rate $H(z)$ and the deviation from Einstein gravity $\eta$ by combining in a model-independent way several observational probes, from redshift distortions, to lensing, to matter clustering.[-]