The solution of the inverse kinematics problem in multi-degrees of freedom robots has been tackled, through the last three decades, by several different approaches including analytical, geometrical, differential and numerical methods. All these techniques present their own advantages and drawbacks. However, a guideline on which approach is better to follow, depending on the kind of task to perform and the type of robotic device used, is still missing. In this work, a quantitative comparative analysis of three different inverse kinematics methodologies for the control of rehabilitative robotic devices is proposed, with aim of devising best practices and guidelines for the selection of the most suitable approach. The analyzed methodologies are implemented and numerically tested on two actual devices, specifically an upper-limb exoskeleton and an upper-limb prosthetic arm.