The mechanical response of plate-based wood furniture is governed by the elastic behavior of the panels and the stiffness of the connections between components. In this work, we propose a combined experimental--computational strategy to (i) identify the transversely isotropic elastic properties of particleboard and (ii) characterize the bending stiffness of L-shaped corner joints. Three-point bending tests on particleboard specimens and bending resistance tests on screw- and dowel-type joints were performed and monitored by Digital Image Correlation (DIC), providing full-field 2D displacement measurements. Particleboard elastic properties were identified through a two-step approach relying on a Timoshenko beam model and a 2D finite element model of the bending test. The experimentally observed variability is represented through probabilistic models for (i) the random elasticity tensor of particleboard within the transversely isotropic symmetry class and (ii) the random joint bending stiffness of screw- and dowel-type connections. Both stochastic models are formulated within a maximum entropy (MaxEnt) framework. The hyperparameters governing the random elasticity tensor are estimated using a least-squares approach combined with Metropolis-Hastings Markov chain Monte Carlo (MCMC) sampling, while the hyperparameters associated with the joint bending stiffnesses are estimated by maximum likelihood using the available experimental dataset. Finally, standardized validation tests on plate-based furniture under static loading are simulated by Monte Carlo sampling to predict the stochastic structural response. Numerical predictions agree well with experimental measurements for all validation configurations considered, demonstrating the relevance of the proposed uncertainty quantification and inverse identification framework for virtual testing of wood-based furniture.