1. The problem statement, all variables and given/known data I was trying to fit experimental data which are theoretically modeled by an exponential function. They are shown in the first graph below. For the first set (red line), the data does not appear to be exponential. My lecturer once said that I need to flip the signs of the first few data points to get an exponential (see the second plot). Is there any justification for doing this? Are there any situations where such a thing would be appropriate? This data relate to the absolute values of magnetisation in an MRI experiment. 2. Relevant equations 3. The attempt at a solution When I flipped the signs of the first six data points, I got a curve which could then be modeled with an exponential, just like the other data sets. I got R2 = 0.99971, and the time constant I calculated from this was in agreement with the expected value. So, why does this method work? Could there be any justification for doing this? Any help would be greatly appreciated.