The number of Internet of Things (IoT) applications and devices has been continuously increasing and is expected to continue to do so in the upcoming years. In this context, we study the uplink uncoordinated spectrum and power allocation problems in a non-orthogonal multiple access (NOMA) IoT network while minimizing the age of information (AoI). Power-limited IoT devices aim at auto-organizing their transmissions on available resources, i.e., the subband-power level pairs, without any coordination between them. To enhance spectral efficiency, full-duplex relays are deployed with queuing capabilities. To solve the problem of minimizing the AoI in both the coordinated and uncoordinated communication settings, we propose algorithms based on matching theory and the multi-armed bandit (MAB) framework respectively. Simulation results show that the proposed uncoordinated MAB-based solution achieves full rate satisfaction for the users. Moreover, it reveals to be very efficient in minimizing the system AoI, transmit power and overhead compared to the coordinated solution.