In this methods paper, we report on the design of a clinical study testing the efficacy of a newly developed control scheme for robot-aided rehabilitation. To measure the value added by a new control scheme, we pursued a parallel-group controlled clinical study design. This approach enables comparing the effects of the novel scheme, based on the Assist-As-Needed (AAN) paradigm, with those of a less sophisticated, fixed gain, Subject-Triggered (ST) controller. We describe the steps followed in the design of this clinical study, including details on the implementation of the two control modes, and a power analysis to determine the required number of subjects to test a clinically significant difference hypothesis. Finally, we present a method for sequential group assignment with co-variates minimization, capable of guaranteeing a desired level of balance of prognostic factors in the two study groups, a crucial requisite for small-scale clinical studies in rehabilitation. To the best of our knowledge, the study presented is the first one testing, in a controlled fashion, the differential effects of a specific control mode in upper extremity rehabilitation after incomplete spinal cord injury.