# Introduction

The mean and standard deviation of a set of data describe the central tendency (the average value) and the spread of a set of data.

# References

• Wikipedia, Mean and Standard deviation
• John E. Freund, Mathematical Statistics, 5th Ed., Prentice Hall, 1992.

# Theory

Suppose we want to describe a set of numeric values with a single number. There are a number of possible descriptions, such as the mean, median, or mode. From high school, you have probably learned that these are the arithmetic average, the middle value, and the most common value.

For reasons that are beyond the scope of this course, we would like to describe data by that value which minimizes the sum of the squares of the differences (or errors):

From calculus, to find this value, all we need do is differenate it with respect to the parameter μ, equate to 0, and solve for μ.

Using the properties of differentiation (the derivative of a finite sum is the sum of the derivatives), we get:

The left-hand sum, however, simplifies to and we may therefore divide both sides by n to get:

Thus, the arithemtic mean is that value which minimizes the sum of the squares of the errors. If we use any other value in our formula for SSE, it must increase the sum.

Now that we have calculated the mean and the SSE, what is the average squared error for each entry? We will call this value σ2 and call it the variation (or variance) of the data:

We may approximate the average error by taking the square root of both sides to yield the standard deviation:

The standard deviation gives a very good bound on how far away the data is from the mean. In the most general case, the Chebyshev inequality states that at least (1 − 1/k2)×100% of the data points fall within k standard deviations from the mean. For example:

• 50% of all data falls within the interval [μ − √2σ, μ + √2σ] (within 1.414 standard deviations),
• 75% of all data falls within the interval [μ − 2σ, μ + 2σ] (within 2 standard deviations),
• 88.9% of all data falls within the interval [μ − 3σ, μ + 3σ] (within 3 standard deviations).

If the data, however, can be said to come from a normal distribution (a bell curve or Gaussian distribution), then it is possible to be much more precises as to the interval:

• 50% of all data falls within 0.6745 standard deviations: [μ − 0.6745σ, μ + 0.6745σ],
• 75% of all data falls within 1.105 standard deviations: [μ − 1.105σ, μ + 1.105σ],
• 88.9% of all data falls within 1.593 standard deviations: [μ − 1.593σ, μ + 1.593σ],
• 95% of all data falls within 1.960 standard deviations: [μ − 1.960σ, μ + 1.960σ],

This last value, 95%, is more often reported as nineteen times out of twenty in polls and other surveys.

# Problem

Given a set of data x1, ..., xn, describe the average value and the spread.

# Assumptions

No assumptions are made on the data.

# The Mean

Calculate:

μ = (x1 + ··· + xn)/n

Calculate:

# Applications to Engineering

Statistics describe reality and many natural phenomena have a normal distribution. Consequently, the mean and standard deviation well define reality.