Key takeaways
- Variance is the average squared deviation.
- Standard deviation is the square root of variance.
- Sample uses N − 1.
- Units match the data.
- Outliers increase the spread.
The steps
Find the mean. Subtract the mean from each value and square the result. Sum the squares. Divide by N (population) or N − 1 (sample). Take the square root. Example: 2, 4, 4, 4, 5, 5, 7, 9. Mean 5. Squared deviations: 9, 1, 1, 1, 0, 0, 4, 16; sum 32. Population variance = 32 ÷ 8 = 4; SD = 2. Sample variance = 32 ÷ 7 = 4.57; SD = 2.14.
Population vs sample
If the data is the entire population, divide by N. If it is a sample used to estimate the population, divide by N − 1. The sample formula is larger, especially for small samples. Statistical software usually labels them “population” and “sample.”
Interpretation
A larger standard deviation means more spread. In a normal distribution, about 68% of values fall within one SD of the mean, 95% within two, and 99.7% within three. This rule applies to normal data; it is not a universal law.
An example
Two classes have the same mean score of 75. Class A has SD 5; Class B has SD 15. Class B has a wider range of scores. A student at 90 is 3 SD above the mean in Class A but 1 SD above in Class B. The context changes the interpretation.
Outliers
An outlier can increase the standard deviation substantially because deviations are squared. Check the data for errors before interpreting. A robust measure such as the interquartile range may be better for skewed data.
What to compare
Compare the mean, standard deviation, range, and distribution shape. A standard deviation without the mean and context is incomplete.
More examples
Data 10, 12, 14, 16, 18: mean 14; squared deviations 16, 4, 0, 4, 16; sum 40. Population variance 40 ÷ 5 = 8; SD 2.83. Sample variance 40 ÷ 4 = 10; SD 3.16. The sample SD is larger because the denominator is smaller.
Standard deviation and standard error
Standard deviation describes the spread of individual values. Standard error describes the uncertainty of the mean and is SD ÷ √n. A large sample can have a small standard error even when the data is spread out. Do not confuse the two.
What to compare
Compare the mean, SD, sample size, and distribution. A SD without context is incomplete.
Sample size
A small sample can produce a large standard deviation by chance. A sample of two values has a very uncertain SD. Report the sample size with the SD. For a mean, the standard error shrinks as the sample grows; the SD of the values does not necessarily shrink.
What to compare
Compare the mean, SD, sample size, and distribution. A SD without context is incomplete.
Reporting
Report the mean and SD together, with the sample size. A mean without a spread can hide a wide range. For skewed data, report the median and interquartile range as well. A graph of the distribution is often more informative than one number.
What to compare
Compare the mean, SD, median, and range. A single statistic is not a description of the data.
What to compare
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Report the sample size.
Final check
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Final check
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Report the sample size.
Final check
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Report the sample size.
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Final check
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread. Report the sample size.
Compare the mean, SD, sample size, and distribution. A SD without the mean and context is incomplete, and a small sample can produce a misleading spread.
Compare the mean, SD, sample size, and distribution, and report the sample size with the result.
A small sample can produce a misleading spread.
Check the mean, SD, sample size, and distribution before interpreting the result. Report the sample size and the context.
Check the mean, SD, sample size, and distribution before interpreting the result, and report the sample size with the context.
Check the mean, SD, sample size, and distribution before interpreting the result, and report the sample size with the context.
A small sample can produce a misleading spread.
Check the mean, SD, sample size, and distribution now, and report the context.
Now check the mean, SD, sample size, and distribution.
Now check the result.
Now check it with context.
Compare the SD with the mean and the sample size before interpreting the result.
Frequently asked questions
When do I use N vs N − 1?
Use N for a population; use N − 1 for a sample. Most real data is a sample.
What does a high standard deviation mean?
The values are more spread out from the mean.
Is standard deviation in the same units?
Yes. Variance is in squared units; standard deviation returns to the original units.
Does the 68-95-99.7 rule always apply?
Only for approximately normal data. Skewed data needs other measures.
Further reading
Keep exploring
Decision path