Variance and Standard Deviation Units

Why Do Spread Measures Have Different Units?

Suppose a set of measurements is recorded in centimeters. The distance between an observation and the mean is also measured in centimeters. But when those distances are squared to calculate variance, the units are squared too: centimeters squared. Taking the square root to obtain standard deviation returns the measurement to centimeters.

This difference is important when reporting results. Variance is useful as a mathematical measure of spread, but its squared units can be awkward to interpret as a distance in the original context. Standard deviation is often easier to describe because it uses the same units as the observations. Neither measure is unit-free, and neither should be reported without its units.

Definition: If observations are measured in units \(U\), their deviations from the mean are measured in \(U\), and their squared deviations are measured in \(U^2\). Therefore, variance is measured in squared units \(U^2\). Standard deviation is the square root of variance, so it is measured in the original units \(U\).

Where the Units Come From

For a population of \(N\) observations \(x_1,\ldots,x_N\), the population variance is the mean of their squared deviations from the population mean \(\mu\). For a sample of \(n\) observations, the sample variance uses \(n-1\) in the denominator. The standard deviation is the square root of the corresponding variance.

\sigma^2=\frac{\sum_{i=1}^{N}(x_i-\mu)^2}{N} \qquad s^2=\frac{\sum_{i=1}^{n}(x_i-\bar{x})^2}{n-1}

\sigma=\sqrt{\sigma^2} \qquad s=\sqrt{s^2}

To track the units, suppose every \(x_i\) is measured in units \(U\). The mean has units \(U\), so subtracting the mean from an observation gives a deviation in \(U\). Squaring that deviation gives units \(U^2\). The denominators \(N\) and \(n-1\) are counts, not measurements, so dividing by them does not change those units. Thus, either variance has units \(U^2\). Taking a square root gives units \(\sqrt{U^2}=U\), which are the original measurement units.

Key distinction: If the data are measured in meters, variance is in square meters (\(\mathrm{m}^2\)) and standard deviation is in meters (\(\mathrm{m}\)). If the data are measured in seconds, variance is in square seconds (\(\mathrm{s}^2\)) and standard deviation is in seconds (\(\mathrm{s}\)).

Worked Measurements

Worked Example: Variability in Sample Lengths

Suppose, hypothetically, that four lengths are measured in centimeters: \(10, 12, 14,\) and \(16\) cm. Treat these four observations as a sample. Find the sample variance and sample standard deviation, including their units.

First find the sample mean:

\bar{x}=\frac{10+12+14+16}{4} =\frac{52}{4} =13\text{ cm}

The deviations from the mean are \(-3,-1,1,\) and \(3\) cm. Squaring them gives \(9,1,1,\) and \(9\) \(\mathrm{cm}^2\), with a sum of \(20\) \(\mathrm{cm}^2\). Since these measurements form a sample, divide by \(n-1=3\):

s^2=\frac{20\text{ cm}^2}{4-1} =\frac{20}{3}\text{ cm}^2 \approx 6.667\text{ cm}^2

Now take the square root to obtain the sample standard deviation:

s=\sqrt{6.667\text{ cm}^2} \approx 2.582\text{ cm}

The variance is about \(6.667\) square centimeters, while the standard deviation is about \(2.582\) centimeters. The standard deviation is on the same scale of units as the measured lengths; the variance is on a squared scale.

Worked Example: When the Measured Quantity Is Already an Area

Suppose, hypothetically, that three small plots have areas of \(8, 10,\) and \(12\) square meters. Treat these as a sample of plots. Find the variance and standard deviation, and check the units carefully.

The sample mean area is:

\bar{x}=\frac{8+10+12}{3}=10\text{ m}^2

The deviations are \(-2,0,\) and \(2\) \(\mathrm{m}^2\). Their squared values are \(4,0,\) and \(4\) \(\mathrm{m}^4\). The sum is \(8\) \(\mathrm{m}^4\), so the sample variance and standard deviation are:

s^2=\frac{8\text{ m}^4}{3-1} =4\text{ m}^4 \qquad s=\sqrt{4\text{ m}^4} =2\text{ m}^2

Here the original measurement is area, so its unit is already square meters. The standard deviation is \(2\) square meters, not \(2\) meters. The variance is in the square of the original unit, \((\mathrm{m}^2)^2=\mathrm{m}^4\). The general rule is that standard deviation returns to the units of whatever was measured, even when those units are themselves squared or otherwise named.

Why Standard Deviation Is Often Easier to Interpret

Variance records spread using squared deviations. Squaring makes every contribution nonnegative and gives greater weight to larger deviations, but it also moves the summary onto a squared measurement scale. If a variable records time in minutes, a variance of \(9\) has units of \(\mathrm{minutes}^2\), not minutes. It does not mean that observations are typically \(9\) minutes from the mean.

Standard deviation is the square root of variance, so it gives a spread measure in the original units. If a set of waiting times has a standard deviation of \(3\) minutes, it is reasonable to describe the standard deviation as a measure of the times’ spread on a scale of a few minutes. That wording does not claim that every waiting time is exactly \(3\) minutes from the mean, or that \(3\) minutes is the average absolute distance. The standard deviation is based on squared deviations, not absolute deviations.

A useful way to keep the roles straight is to ask two questions: “What units were the observations measured in?” and “Was the summary formed before or after taking a square root?” Variance has squared units because it is based on squared deviations. Standard deviation has the original units because the square root reverses the squaring of the unit.

Unit Conversion and Squared Units

A unit conversion applies to variance differently from how it applies to standard deviation. For example, \(1\) centimeter is \(10\) millimeters. Since variance uses squared measurements, one square centimeter equals \(10^2=100\) square millimeters. Standard deviation, by contrast, is converted using the original length conversion: multiply centimeters by \(10\) to express the result in millimeters.

Worked Example: Converting Variance and Standard Deviation

Use the sample of lengths \(10,12,14,\) and \(16\) cm from the earlier calculation. Its sample variance is \(6.667\) \(\mathrm{cm}^2\), and its sample standard deviation is \(2.582\) cm. Express both summaries in millimeters and square millimeters.

Because \(1\text{ cm}=10\text{ mm}\), square both sides of the unit conversion to convert variance:

1\text{ cm}^2=(10\text{ mm})^2 =100\text{ mm}^2

Therefore, the variance in square millimeters is:

6.667\text{ cm}^2 \times \frac{100\text{ mm}^2}{1\text{ cm}^2} =666.7\text{ mm}^2

The standard deviation is converted as an ordinary length:

2.582\text{ cm} \times \frac{10\text{ mm}}{1\text{ cm}} =25.82\text{ mm}

The numerical values change by different factors because their units are different: the variance is in squared units, while the standard deviation is in the original units. These are the same summaries expressed on different measurement scales, not different amounts of variability.

A Short Justification of the Unit Rule

The rule follows directly from the calculation, rather than from a special convention about naming units. Let the measurement unit be \(U\). Every deviation \(x_i-\bar{x}\) has unit \(U\), and its square has unit \(U\times U=U^2\). A variance is a sum of such squared deviations divided by a count, so the variance retains unit \(U^2\). The standard deviation is defined as the nonnegative square root of that variance. Since \(\sqrt{U^2}=U\) for a positive measurement unit, its unit is \(U\).

For a sample, the divisor \(n-1\) does not affect the unit argument: \(n-1\) counts observations and carries no measurement unit. The same is true of the population divisor \(N\). Thus, the distinction between the sample and population formulas affects the numerical calculation, but not the conclusion about units.

Common Mistakes

Calling variance a typical distance. The wrong reasoning is: “The variance is \(6.667\), so observations are typically \(6.667\) cm from the mean.” The variance in the length example is \(6.667\) \(\mathrm{cm}^2\), not centimeters. Its square root, about \(2.582\) cm, is the standard deviation.
Giving standard deviation squared units. The wrong reasoning is: “The calculation involved squared deviations, so standard deviation is in square centimeters.” The square root returns the units to centimeters.
Converting variance like a length. The wrong reasoning is to multiply a variance in \(\mathrm{cm}^2\) by \(10\) when converting to \(\mathrm{mm}^2\). A square-centimeter conversion uses the squared length factor: \(1\text{ cm}^2=100\text{ mm}^2\).
Assuming a standard deviation is an average absolute distance. Standard deviation is based on squared deviations and a square root. It is not calculated by averaging the absolute deviations from the mean.

Key takeaway: Variance has the square of the data’s units because it is calculated from squared deviations. Standard deviation takes the square root of variance and therefore uses the original measurement units. Always include the units when reporting either measure.

Check Your Understanding

Use the unit rules to answer each question.

A sample of distances is recorded in meters. What are the units of its sample variance and sample standard deviation?
A variance for waiting times is reported as \(16\) \(\mathrm{minutes}^2\). What is the corresponding standard deviation, including its units?
Areas are recorded in square centimeters. What units should be used for their variance and standard deviation?
A variance is \(3\) \(\mathrm{cm}^2\). A student converts it to \(30\) \(\mathrm{mm}^2\). Identify the error and give the correct converted value.
Why is it inaccurate to describe a variance of \(25\) \(\mathrm{seconds}^2\) as a typical distance of \(25\) seconds from the mean?