Cognitive radios (CRs) can dynamically reconfigure their transmission and/or reception parameters. In a cognitive radio ad hoc network (CRAHN) setting, such reconfiguration is challenging due to the lack of centralized control and fixed infrastructure. In this paper, we propose a method to improve the throughput of secondary users (SUs) in a CRAHN by dynamically adapting their sensing and transmission times. First, we conduct a designed experiment on five CR parameters in the ns-2 network simulator with extensions to support CRAHNs. A statistical analysis of the resulting data attributes the contribution of each parameter, and interactions among them, to SU throughput. Based on these results, we propose for each SU to measure its instantaneous throughput and adapt its sensing and transmission times according to the predicted throughput. Simulation results demonstrate that SUs achieve higher throughput by adapting their sensing and transmission times than by using default static values for these parameters.