<p>Detecting and quantifying phase synchronization of chaotic oscillators are not trivial tasks. Many approaches involve creating a phase model, or estimating the phase. An approach is here introduced that does not require a phase definition, rather it is based on the monitoring of the so-called latent variables, or principal components (PCs). Although PCs are well-known in signal processing, their use in synchronization problems has not been fully exploited. As the phase synchronization deteriorates, the latent space becomes disorganized and this can be characterized through a rather simple measure taken from principal components analysis, namely the <i>explained variance ratio</i> (EVR). It is shown that such a measure of the latent space correlates well with an already established phase synchronization measure, the mean phase coherence. Further, by monitoring this ratio when increasing the coupling strength in a network, it is possible to identify the formation of clusters of synchronized oscillators. The proposed methodology has shown to be consistent over a range of scenarios, including pairs of coupled oscillators, and different topologies of networks. The EVR can be used in cases where phase estimation is difficult or inaccurate or in larger networks, where the phase error has to be computed pairwise. It is shown that EVR can also be a measure of generalized synchronization of systems with different topologies.</p>