How Adding or Scaling Data Changes the Mean and Standard Deviation

Many statistics questions describe a data set after every observation has been adjusted. Instead of recomputing a mean and standard deviation from scratch, identify whether the operation shifts the values or rescales them. A shift affects location; a scale affects both location and spread.

Add or subtract the same number

If each value x becomes x + c, the mean and median each increase by c. The minimum, maximum, quartiles, and mode also shift by c when applicable. The range, interquartile range, variance, and standard deviation do not change because every pairwise distance stays the same.

Multiply every value by a number

If each value becomes ax, measures of center are multiplied by a. Measures of spread use the size of the scale: range, interquartile range, and standard deviation are multiplied by |a|, while variance is multiplied by a². If a is negative, the order reverses, so the new minimum comes from the old maximum, but the spread remains nonnegative.

Worked example

Suppose a class has mean score 70 and standard deviation 8. If every score receives five bonus points, the new mean is 75 and the standard deviation remains 8. If instead every score is doubled, the mean becomes 140 and the standard deviation becomes 16; the variance becomes four times its original value.

Units explain the rule

Adding five minutes to every recorded time changes the center but not how far apart observations are. Converting a measurement from meters to centimeters multiplies every value by 100, so the center and standard deviation also scale by 100. Variance uses squared units, which is why its scale factor is squared.

Common errors

• Changing standard deviation when a constant is added.
• Multiplying variance by a instead of a².
• Using a negative spread after multiplying by a negative number.
• Confusing a shift of every observation with adding a new observation to the data set.

Key takeaway

A translation changes center but preserves spread. A scale changes both: multiply standard deviation by the absolute scale factor and variance by its square.

Separate shifts from rescaling

Adding the same constant to every observation shifts the entire distribution without changing its shape or spread. If every score rises by 4 points, the mean and median rise by 4, and the range, interquartile range, and standard deviation stay the same. Multiplying each observation by a positive constant changes both location and scale: multiplying scores by 3 triples the mean, median, range, and standard deviation. Dividing by 3 divides each of those measures by 3. A negative multiplier also reverses the order of values; standard deviation and range are multiplied by its absolute value, not made negative.

Consider the data 2, 4, and 6. Their mean is 4 and their population standard deviation is about 1.63. Adding 10 produces 12, 14, and 16: the mean becomes 14 while the standard deviation remains about 1.63. Multiplying the original values by 2 produces 4, 8, and 12: the mean becomes 8 and the standard deviation becomes about 3.27. The examples distinguish a change in center from a change in units or spread.

Standardize values with a z-score

A z-score changes a raw value into standard-deviation units: z = (x − μ) / σ for a population, or z = (x − x̄) / s when a sample mean and sample standard deviation are being used. Subtracting the mean centers the data at zero; dividing by the standard deviation rescales the spread to one. The transformation preserves relative position: a value two standard deviations above its original mean receives z = 2. A negative z-score means the value is below the mean.

For scores with mean 70 and standard deviation 8, a score of 86 has z = (86 − 70) / 8 = 2. This says it is two standard deviations above the group mean. It does not say the score is in the 98th percentile unless a distribution model is also given. A normal-distribution percentile interpretation requires the normality assumption; the arithmetic z-score itself does not.

Watch for a change of units

Converting measurements from meters to centimeters multiplies every value by 100. The mean and standard deviation are therefore multiplied by 100, while z-scores remain unchanged because both the deviation from the mean and the standard deviation receive the same factor. Converting temperatures by adding an offset and multiplying by a scale factor changes location and scale differently. A pure additive offset leaves standard deviation unchanged; the scale factor changes it by its absolute value.

• If 5 is added to every value, add 5 to mean and median; do not change standard deviation.
• If every value is doubled, double the mean and standard deviation.
• If every value is multiplied by −2, reverse the order and double spread measures.
• Do not claim a z-score gives a percentile without a distribution model.
• Use the same data set and same standard-deviation convention when comparing calculations.

Transform the range and variance consistently

If every value is multiplied by a constant c, the range is multiplied by |c| and the variance by c²; standard deviation is multiplied by |c|. Multiplying by −2 reverses order and doubles distances between observations, so range and standard deviation double while variance quadruples. These formulas follow from how deviations from the transformed mean scale.

A useful check is to transform a simple set such as 1, 3, 5. Adding 4 gives 5, 7, 9: the mean increases by 4, but every pairwise distance is unchanged. Multiplying by 3 gives 3, 9, 15: the mean and standard deviation scale by 3, while variance scales by 9. Keep units in mind because variance uses squared units.

Apply the distinction carefully

Adding a constant to every observation shifts the mean, median, and other location measures by that constant, while leaving range and standard deviation unchanged. Multiplying every value by a positive constant multiplies location measures and measures of spread by that factor. For example, converting a set of dollar amounts to cents scales each observation by 100, so its mean and standard deviation also scale by 100, while the shape of the distribution stays the same. If the multiplier is negative, order reverses and the median changes sign, while measures of spread scale by the absolute value. These rules help check calculations and interpret unit conversions; do not confuse a change in units with a change in the underlying pattern.