Normal Variables and Their Parameters

Why Normal Variables Are Useful

Many measurements vary around a typical value: repeated readings from an instrument, adult heights in a defined population, or the dimensions of parts made by a stable process. A normal model is useful when such variation is reasonably concentrated around a center, tapers away from that center in both directions, and has an approximately symmetric, bell-shaped pattern.

Calling a measurement normal is a statement about a model for its variation, not a claim that every observed value equals the average or that real data must follow a perfect curve. The mean and standard deviation specify the model’s location and scale. Understanding those parameters makes it possible to describe the model before using it for probability calculations, which are developed in later material.

Definition: A random variable \(X\) is normally distributed with mean \(\mu\) and standard deviation \(\sigma\) if its distribution is a normal distribution with those parameters. The notation is \(X\sim N(\mu,\sigma^2)\), where \(\mu\) is the mean, \(\sigma^2\) is the variance, and \(\sigma>0\) is the standard deviation. The normal model is continuous, symmetric about \(\mu\), and bell-shaped.

The two numbers in \(N(\mu,\sigma^2)\) have different roles and different units. The mean \(\mu\) is in the same units as \(X\); it locates the model’s center. The variance \(\sigma^2\) is in squared units. Its positive square root, \(\sigma\), is the standard deviation and is in the same units as \(X\). For example, if \(X\) is a length in centimeters, its mean and standard deviation are in centimeters, but its variance is in square centimeters.

The Center: Mean \(\mu\)

The mean is the balance point of the distribution. In a normal model, the curve is symmetric around \(\mu\): values equally far below and above \(\mu\) occupy matching positions in the shape. The mean is also the median, the value that divides the distribution into equal halves, and the mode, the value at the peak of the bell. These coincide because the normal distribution is symmetric and has one central peak.

The mean describes a model’s center, not a prediction that a particular observation will equal \(\mu\). Individual measurements vary around it. Nor should a population or model mean be confused with the sample mean \(\bar{x}\), which is calculated from observed data. A sample mean may be used to estimate a model mean, but it is not automatically the same quantity.

Worked Example: Interpreting the Center of a Measurement Model

Suppose, hypothetically, that the lengths of a certain type of manufactured component are reasonably modeled by \(X\sim N(120,16)\), where measurements are in millimeters. Interpret the model’s center and identify its standard deviation.

In the notation \(N(\mu,\sigma^2)\), the first parameter is the mean and the second is the variance. Therefore,

\mu=120\text{ mm},\qquad \sigma^2=16\text{ mm}^2

The standard deviation is the positive square root of the variance:

\sigma=\sqrt{16\text{ mm}^2}=4\text{ mm}

Thus the model is centered at 120 mm, and its standard deviation is 4 mm. The symmetry means that the model has the same shape on either side of 120 mm. It does not mean every component is 120 mm long; individual lengths vary around that center.

The Spread: Standard Deviation \(\sigma\)

The standard deviation measures the typical scale of variation around the mean. A smaller \(\sigma\) gives a more concentrated normal shape near \(\mu\); a larger \(\sigma\) gives a more spread-out shape. The center alone does not tell us how much observations vary, and the standard deviation alone does not tell us where the model is centered.

Variance and standard deviation encode the same spread information in different units. By definition, the variance is the expected squared deviation from the mean, \(\operatorname{Var}(X)=E[(X-\mu)^2]\), and the standard deviation is its positive square root, \(\sigma=\sqrt{\operatorname{Var}(X)}\). Squaring a deviation squares its units; taking the square root returns the units to those of the measurement. For example, a variance of \(25\text{ seconds}^2\) corresponds to a standard deviation of \(5\text{ seconds}\), not 25 seconds.

Standard deviation is not the full range of possible values, nor is it a guarantee that every observation lies within a fixed distance of the mean. The normal model has tails extending in both directions. How much probability lies in a particular interval is a separate question; the parameter \(\sigma\) by itself is a measure of scale, not an interval probability.

Worked Example: Comparing Spread While Holding the Center Fixed

Two hypothetical production lines make the same type of part. Suppose Line A’s measurement \(X_A\) is modeled by \(N(50,9)\), and Line B’s measurement \(X_B\) by \(N(50,36)\), with all measurements in millimeters. Compare the centers and spreads.

For Line A, the mean is \(\mu_A=50\) mm and the variance is \(\sigma_A^2=9\text{ mm}^2\), so

\sigma_A=\sqrt{9\text{ mm}^2}=3\text{ mm}.

For Line B, the mean is \(\mu_B=50\) mm and the variance is \(\sigma_B^2=36\text{ mm}^2\), so

\sigma_B=\sqrt{36\text{ mm}^2}=6\text{ mm}.

Both models are centered at 50 mm. Line B has twice the standard deviation of Line A, since \(6/3=2\), so its normal shape is more spread out around the same center. This comparison describes the models; it does not establish that either line’s real measurements follow its assumed model.

The Bell-Shaped Form and What the Parameters Do

A normal distribution has one peak at its center and is symmetric: moving the same distance to the left or right of \(\mu\) gives corresponding positions in the shape. The shape tapers toward both tails rather than stopping at a finite minimum or maximum. The parameters control two distinct features: changing \(\mu\) moves the center, while changing \(\sigma\) changes the spread.

These features provide a useful way to compare models. If two normal variables have equal standard deviations but different means, their shapes have the same spread and are centered at different values. If they have the same mean but different standard deviations, their centers match but their spreads differ. The variance parameter in the notation is \(\sigma^2\), so the standard deviation must be obtained by taking its positive square root before describing the spread in the original measurement units.

Parameter change	What stays the same	What changes in the normal model
Change \(\mu\), keep \(\sigma\) fixed	Spread and overall shape	The center moves to the new mean
Change \(\sigma\), keep \(\mu\) fixed	Center	The spread changes; larger \(\sigma\) means more spread
Change both \(\mu\) and \(\sigma\)	Symmetric bell-shaped form	Both center and spread change

The symmetry also explains why the normal model’s mean is its center. For every location \(\mu-d\) at distance \(d\) below the center, there is a matching location \(\mu+d\) at distance \(d\) above it. Their deviations from \(\mu\) are \(-d\) and \(+d\), which cancel when paired. Since a normal distribution is symmetric and has a finite mean, these paired deviations balance, leaving the mean at \(\mu\). Its single peak is at that same central value, so the mode is \(\mu\) as well.

Worked Example: Translating a Description into Parameters

Imagine a laboratory instrument whose repeated readings, in hypothetical conditions, are modeled as normally distributed. The model is described as having a mean of 2.40 volts and a variance of \(0.0100\text{ volts}^2\). Write its normal notation and explain the center and spread.

The first parameter is the mean, \(\mu=2.40\) volts. The second parameter in normal notation is the variance, \(\sigma^2=0.0100\text{ volts}^2\). Taking the positive square root gives

\sigma=\sqrt{0.0100\text{ volts}^2}=0.100\text{ volts}.

Therefore, the model is

X\sim N(2.40,\ 0.0100),

with voltage as the measurement unit. The model is centered at 2.40 volts and has a standard deviation of 0.100 volts. Its shape is symmetric and bell-shaped around 2.40 volts. The variance is reported in squared volts, whereas the standard deviation uses volts, the same unit as each reading.

When Is a Normal Model Plausible?

A normal model is an assumption about a variable’s distribution. It can be a useful approximation for measurements influenced by many small sources of variation, particularly when the resulting pattern is roughly symmetric and has one central peak. But the context and observed pattern matter. Strong skewness, multiple peaks, unusual extreme values, or a hard physical boundary near the center may make a normal model inappropriate or require careful qualification.

A normal distribution extends over all real values. This can be unrealistic for a variable that cannot be negative, such as a length or a duration. A normal model may still be a reasonable approximation if its relevant range is far from the boundary, but the possibility of impossible model values should be considered. Choosing a mean and standard deviation does not, by itself, establish that the model fits; the shape of the measurement process and data should support the assumption.

For example, an approximately symmetric distribution of adult heights within a carefully defined group may plausibly be modeled as normal. A waiting time with many short waits and a smaller number of very long waits is often right-skewed, so a normal model may not represent it well. The issue is not whether a measurement is numerical, but whether its distribution is adequately described by the normal model.

Common Mistakes

Reading the second normal parameter as the standard deviation. In \(N(\mu,\sigma^2)\), the second entry is the variance. Take its positive square root to obtain \(\sigma\).
Mixing up units. If measurements are in centimeters, \(\mu\) and \(\sigma\) are in centimeters, but \(\sigma^2\) is in square centimeters.
Treating the mean as the value of every observation. The mean is the model’s center; individual measurements vary around it.
Assuming that a normal model is automatically suitable. The model’s symmetric, single-peaked shape should be plausible for the measurement context.

Key takeaway: In \(X\sim N(\mu,\sigma^2)\), \(\mu\) gives the center, while \(\sigma=\sqrt{\sigma^2}\) gives the spread in the measurement’s units. A normal variable has a symmetric, bell-shaped distribution, but whether that model suits a real measurement must be judged from the context and pattern of variation.

Check Your Understanding

Try these questions without looking back.

For \(X\sim N(18,25)\), identify the mean, variance, and standard deviation. If \(X\) measures time in minutes, state the units of each parameter.
Two normal models have the same standard deviation but different means. Which feature of their shapes differs, and which feature is shared?
A normal model has variance \(0.0400\text{ kg}^2\). Find its standard deviation and give the units.
Why does the mean of a normal distribution coincide with its center of symmetry?
Give one feature of a real measurement pattern that could make a normal model questionable, and explain why.