I'm not going to open the file. You say "It is the measure of 4 different treatments" so this is just numerical data? Numerical derivatives are notoriously inaccurate- they involve the division of two numbers each close to 0. The simplest, but least accurate, method is to just subtract two consecutive function values, then divide by the difference in the independent variables. My personal preference is to use all of the data given to construct a polynomial function that gives those values, then differentiate that polynomial.