Tutorials › Introduction to Statistics › Avoiding Sign and Unit Errors in Summary Calculations

Home  /  Tutorials  /  Center, Spread and Position

Center, Spread and Position · Tutorial 174 of 996

Avoiding Sign and Unit Errors in Summary Calculations

Use sign and unit checks to calculate and interpret common summary statistics without confusing deviations, squared deviations, standard deviations, or z scores.

Beginner 9 min read

What You'll Learn

  • Distinguish signed deviations from nonnegative squared deviations.
  • Check that deviations from the sample mean sum to zero.
  • Keep track of the units for deviations, variance, and standard deviation.
  • Calculate sample standard deviation only when the sample size is at least two.
  • Compute z scores using compatible units and interpret their signs correctly.
  • Recognize when zero spread makes a z score undefined.

Why Sign and Unit Checks Matter

A summary calculation can look convincing and still be wrong. A missed negative sign can make an observation appear to be above the mean when it is below it. A unit mismatch can change a z score even though the underlying measurement has not changed. Keeping track of signs and units provides simple checks that help catch these errors before interpreting a result.

The basic operations are linked: subtract a center from an observation to find its deviation, square deviations when calculating variance, take a square root to obtain a standard deviation, and divide a deviation by a standard deviation to obtain a z score. At each stage, the sign and units follow predictable rules.

Definition: For an observation \(x_i\) and sample mean \(\bar{x}\), the deviation is \(x_i-\bar{x}\). It is signed: negative values indicate observations below the mean, positive values indicate observations above it, and zero indicates an observation equal to the mean. The squared deviation is \((x_i-\bar{x})^2\), which is always nonnegative.

For a sample of \(n\ge 2\) quantitative observations measured in the same units, the sample standard deviation is

$$ s=\sqrt{\frac{\sum_{i=1}^{n}(x_i-\bar{x})^2}{n-1}}. $$

The numerator is the sum of squared deviations. Dividing it by \(n-1\) gives the sample variance, and taking the square root gives \(s\). For a population, the population variance is the mean of the squared deviations from \(\mu\), and the population standard deviation \(\sigma\) is its square root. These definitions distinguish variance, which uses squared units, from standard deviation, which returns to the original measurement units.

A Quick Sign Check: Deviations Add to Zero

The deviations from the sample mean must sum to zero, apart from small rounding differences if calculations use rounded values. This gives a useful arithmetic check.

$$ \sum_{i=1}^{n}(x_i-\bar{x}) =\sum_{i=1}^{n}x_i-n\bar{x} =\sum_{i=1}^{n}x_i-n\left(\frac{\sum_{i=1}^{n}x_i}{n}\right) =0. $$

The equality follows from the definition of the sample mean. If a list of deviations does not add to zero, check the mean, the subtractions, and the signs. Once the deviations are squared, however, their signs are gone: every squared deviation is zero or positive. A negative squared deviation is impossible.

Worked Example: Checking Deviations and Squared Deviations

A hypothetical sample of five resting pulse rates, measured in beats per minute (bpm), is 62, 68, 74, 80, and 86. Calculate the deviations from the sample mean and the sample standard deviation, then use the signs and units to check the work.

Find the mean. The total is \(62+68+74+80+86=370\) bpm, so \(\bar{x}=370/5=74\) bpm.

Calculate each deviation. Subtract 74 bpm from each observation: \(62-74=-12\) bpm, \(68-74=-6\) bpm, \(74-74=0\) bpm, \(80-74=6\) bpm, and \(86-74=12\) bpm. The first two observations are below the mean, the fourth and fifth are above it, and the observation equal to the mean has zero deviation. As a check, \(-12-6+0+6+12=0\) bpm.

Square the deviations. The squared deviations are \(144, 36, 0, 36,\) and \(144\) bpm\(^2\). Their sum is \(360\) bpm\(^2\). Each is nonnegative, as a square must be.

Calculate the sample standard deviation. Here \(n=5\), so \(n-1=4\):

$$ s=\sqrt{\frac{360\ \text{bpm}^2}{5-1}} =\sqrt{90\ \text{bpm}^2} \approx 9.487\ \text{bpm}. $$

The sample variance is \(90\) bpm\(^2\), while the sample standard deviation is approximately \(9.487\) bpm. The units return to bpm after taking the square root. The standard deviation is not negative: a square root used for standard deviation is the nonnegative root.

Track Units Through Each Operation

Suppose the original observations are measurements in a unit such as centimeters, seconds, or points. Subtracting two measurements in the same unit gives a deviation in that unit. Squaring a deviation squares its unit, so a variance has squared units. Taking the square root returns to the original unit. Finally, dividing one deviation by a standard deviation in the same unit cancels the units, so a z score has no units.

QuantityOperationUnits if observations are measured in \(U\)
DeviationObservation minus center\(U\)
Squared deviationDeviation squared\(U^2\)
VarianceAverage of squared deviations\(U^2\)
Standard deviationSquare root of variance\(U\)
Z scoreDeviation divided by standard deviationNo units

A mismatch occurs when the observation, center, and standard deviation are expressed in different units. Convert them all to one unit before subtracting and dividing. Converting only one of the numbers can give a z score that is numerically wrong even though the division itself is performed correctly.

Worked Example: Fixing a Unit Mismatch in a Z Score

In a hypothetical class, a test time is 90 minutes. The class mean is 1.2 hours and the standard deviation is 0.25 hours. Find the test time's z score.

A z score compares a measurement with a center in standard-deviation units. When a reference distribution has mean \(\mu\) and positive standard deviation \(\sigma\), \(z=(x-\mu)/\sigma\). For this example, convert the observed time to hours so every quantity has the same unit: \(90\) minutes \(=90/60=1.5\) hours.

$$ z=\frac{1.5\ \text{hours}-1.2\ \text{hours}}{0.25\ \text{hours}} =\frac{0.3\ \text{hours}}{0.25\ \text{hours}} =1.20. $$

The z score is \(1.20\), with no units. The observed time is 1.20 standard deviations above the mean. The units cancel because both the deviation and the standard deviation are in hours.

The same result follows if all quantities are converted to minutes: the mean is \(1.2(60)=72\) minutes and the standard deviation is \(0.25(60)=15\) minutes. Then \((90-72)/15=18/15=1.20\). By contrast, subtracting 1.2 hours directly from 90 minutes is invalid: the numbers do not share a unit.

Read the Sign of a Z Score Carefully

For a positive standard deviation, the sign of a z score comes from the deviation in its numerator. A positive z score means the observation is above the center used in the calculation; a negative z score means it is below; zero means it equals the center. The magnitude gives the distance from the center in standard-deviation units. This sign rule requires a positive standard deviation: if the standard deviation is zero, the z score is undefined because its formula divides by zero.

For a sample, a standardized value may be calculated as \(z=(x-\bar{x})/s\) when \(s>0\). For a population distribution, the corresponding expression is \(z=(x-\mu)/\sigma\) when \(\sigma>0\). Be clear about which center and standard deviation are being used, and keep the observation and those quantities in compatible units.

Worked Example: Interpreting the Sign and Handling Zero Spread

A hypothetical set of exam scores is summarized by a mean of 78 points and a standard deviation of 6 points. Find and interpret the z score for a score of 69 points. Then consider what happens if every score in another set is 80 points.

For the score of 69, subtract the mean first. The deviation is \(69-78=-9\) points, so the score is below the mean. Divide by the positive standard deviation:

$$ z=\frac{69-78}{6}=\frac{-9\ \text{points}}{6\ \text{points}}=-1.50. $$

The score is 1.50 standard deviations below the mean. Its negative sign is consistent with the negative deviation. If the score had instead been 87 points, its deviation would be \(87-78=9\) points and its z score would be \(9/6=1.50\), indicating a position 1.50 standard deviations above the mean.

Now suppose all observations in another set equal 80 points. Its center is 80 points and its standard deviation is 0 points. For an observation of 80 points, the proposed z score would be \((80-80)/0=0/0\), which is undefined. It is not a z score of zero: division by zero has no defined value. With no spread, the formula cannot express an observation's distance from the center in standard-deviation units.

Common Errors and How to Avoid Them

  • Reversing a subtraction. The deviation is observation minus center, \(x_i-\bar{x}\), not center minus observation. Reversing the order reverses the sign and changes what “above” or “below” the mean means.
  • Giving a squared deviation a negative sign. A negative deviation becomes nonnegative when squared. For example, \((-6)^2=36\), not \(-36\).
  • Reporting variance in the original units. If observations are in meters, variance is in square meters, while standard deviation is in meters. Taking the square root is what returns the units to the original scale.
  • Mixing units in a z-score calculation. Convert the observation, center, and standard deviation to a common unit before calculating. The final z score is dimensionless.
  • Assuming every standard deviation is positive. For a sample with \(n\ge2\), \(s=0\) if all observations are equal. A z score using a zero standard deviation is undefined, not positive, negative, or zero.
  • Using the sample standard deviation formula with one observation. The formula divides by \(n-1\) and applies only for \(n\ge2\). At \(n=1\), the denominator is zero, so that formula does not define \(s\).

A practical check is to write the units next to intermediate results until the final division. Also check the direction before computing: an observation below its center must have a negative deviation, and, when the standard deviation is positive, a negative z score. These checks do not replace careful arithmetic, but they can reveal a mistake quickly.

Key takeaway: Deviations retain their signs and original units; squared deviations are nonnegative and have squared units; standard deviations are nonnegative and have the original units; and z scores have no units. A z score's sign matches the deviation only when the standard deviation is positive.

Check Your Understanding

Use sign and unit checks as well as arithmetic in your answers.

  1. An observation is 14 cm and the mean is 18 cm. What is its deviation, including its sign and units?
  2. If a deviation is \(-3.5\) seconds, what is its squared deviation and its unit?
  3. For a sample of \(n=6\), the squared deviations sum to \(125\) cm\(^2\). Write the sample standard deviation calculation and state its units.
  4. A measurement is 45 minutes, while its mean and standard deviation are 1 hour and 0.25 hours. What conversion should be made before calculating the z score?
  5. When does a negative deviation give a negative z score, and what happens to the z score if the standard deviation is zero?

Previous

Next

Back to Tutorials