You are trying to determine the melting point of a new material of which you have a large number of samples. For each sample that you measure you find a value close to the actual melting point C but corrupted with a measurement error. We model this with random variables. Mi = c + Ui

Where Mi is the measured value in degree Kelvin and Ui is the occurring random error. It is known that E[Ui]=0 and Var[Ui]=3 for each i and that we may consider the random variables M1, M2...independent. According to Chebyshev's inequality how many samples do you need to measure to be 90 percent sure that the average of the measurements is within half a degree of C?