Some of the data sets contain numbers with several decimal places which can become very confusing in an equation, is it possible to round the numbers in the original data set without severely changing the final outcome of the modeling process?
Yes. The deviation from the ideal quadratic behavior is typically at least a few units, and when this is the case it wouldn't hurt a thing to round off to two or three significant figures.