I have a question about fMRI model fitting. So can I use Rsquared to evaluate the fitness of my model to the timeseries, if Rsquared is low than I can say well, maybe we have a wrong model. I'm not sure how this would actually work though. 

Since recently we have had model fitting problems and we just eyeballed the outcomes and say this is not what we expected and then trace back and find out the model's wrong. If there's a way to tell the fitness of the model statistically that will be great.

Thanks,
Wen