There are several methods for Structural Health Monitoring (SHM). Many of them are based on the analysis of measured vibration data, identified modal parameters and subsequent estimations of structural changes. However, most of these methods underly limitations due to various uncertainties at different steps of the analysis. As an alternative, a statistical approach based on measured data is provided to calculate a damage indicator using a hypothesis test. This approach relies on the covariance-driven Stochastic Subspace Identification (SSI-cov) routines and requires some pre-defined parameters which are not necessarily trivial to determine. To reduce respective uncertainties and subjective intervention by the analyst, this paper systematically analyses several parameters and gives recommendations for parameter choices especially for the application of the statistical fault detection algorithm. This work proposes a novel approach employing a clustering routine for the rank estimation. The results are illustrated by the example of an experimental beam structure.