Experimental methods are necessary to quantify structural damping. Among these methods, modal testing of bladed disks are of particular interest because it provides an efficient experimental characterization of these structures which are critical parts of aeronautic turbojet engines. This work presents techniques to determine the optimal order of subspace state-space identification methods for modal parameter estimation of a realistic fan stage. Indeed, the order of the identification model is unknown during modal analysis requiring techniques to determine which model order should be used for parameter estimation. To investigate these techniques, several methods to estimate the optimal model order are reviewed. A specific focus is proposed on the determination of model order through statistical tests as these techniques have the benefits of providing indicators along with a threshold built over hypothesis testing and a probability of rejection. In order to further analyze model order determination techniques, some of these techniques are evaluated by means of a numerical model of a realistic fan stage, hence making possible to assess the performances of estimation methods close to experimental conditions. In particular, estimation in a context of high modal density as well as model order determination through "M-test" are addressed using this model. After assessing the implemented estimation method over the numerical model, this method is applied over experimental data obtained by performing modal tests over a full-scale composite rotating fan in vacuum condition.