Short Answer
Students often notice that a teacher rounds the mean to one decimal place but the standard deviation to two, or rounds the mean to hundredths while the SD appears with thousandths. This site is a precision-and-rounding reference, not just a calculator, so the short answer is: the mean and the standard deviation are different kinds of quantities, and different standards allow them to be rounded differently. The mean is a central estimate that follows the decimal-place precision of the raw measurements. The standard deviation is a measure of dispersion, and when it is used as an uncertainty, it is often rounded to two significant digits. The apparent inconsistency is usually a convention, not an arithmetic error.
Rule Statement
The core rules can be stated in four parts.
- Descriptive statistics of raw measurements: round the mean to the same decimal place as the raw data. If the data are recorded to 0.1, report the mean to 0.1. This follows the addition and subtraction rule for decimal places.
- Standard deviation in lab reports: round the standard deviation to one extra decimal place beyond the raw data, or to two significant digits if it is serving as an uncertainty. If the data are to 0.1, report the SD to 0.01. If the data are to 0.01, report the SD to 0.001.
- Metrological uncertainty statements: follow the GUM and NIST convention. Round the standard deviation, or standard uncertainty, to at most two significant digits. Then round the mean to the same decimal place as the rounded uncertainty.
- Discipline-specific styles: APA style often uses two decimal places for both mean and SD. Engineering and chemistry may use ASTM E29 or ISO 80000-1 conventions based on significant digits and the reported uncertainty.
Thus, the mean and SD can be rounded differently because they are not the same kind of number. The mean shares the unit and scale of the measurements. The SD describes spread and is often treated as an uncertainty.
Convention Comparison Table
| Context or convention | Mean rounding | Standard deviation rounding | Basis |
|---|---|---|---|
| Introductory chemistry or physics lab | Same decimal place as raw data | One extra decimal place | Common lab pedagogy; preserves dispersion information |
| GUM and NIST uncertainty statement | Same decimal place as SD | Two significant digits | JCGM 100:2008, 7.2.6; NIST TN 1297, 7.2.6 |
| APA style in psychology | Two decimal places | Two decimal places | APA reporting conventions |
| ASTM E29 and ISO 80000-1 | Based on significant digits | Based on significant digits | No separate mean/SD rule; consistent rounding is required |
| Teacher example: mean 12.4, SD 0.16 | 12.4 | 0.16 | One-extra-decimal convention for SD |
Standards Citation
Several authoritative standards explain why mean and standard deviation rounding can differ.
- JCGM 100:2008 (GUM), Clause 7.2.6: The numerical values of the estimate and its standard uncertainty should not be given with an excessive number of digits. It is often sufficient to give the standard uncertainty to at most two significant digits. The estimate should be rounded to the same number of decimal places as the uncertainty.
- NIST TN 1297, Section 7.2.6: NIST adopts the same metrological convention: report the uncertainty to two significant digits and round the measured value to the same decimal place.
- ASTM E29-13a, Sections 6-8: This practice covers significant digits, rounding, and conformance decisions. It does not create a separate statistical rule for the mean, but it requires that rounding be consistent and that it not introduce false precision.
- ISO 80000-1:2009, Clause 7.3.4: Rounding should reflect the uncertainty of the value. Retaining more digits than the uncertainty supports is misleading.
- ISO 5725-1:1994, Clause 4.2: This standard addresses trueness and precision of measurement methods and supports reporting standard deviations with attention to the precision of the measurement process.
GUM 7.2.6: The value of the estimate should be rounded to the same number of decimal places as its standard uncertainty.
Worked Examples
Example 1: Data Recorded to the Tenths
Data: 12.3, 12.5, 12.4, 12.6, 12.2.
- Sum = 62.0; n = 5; mean = 62.0 / 5 = 12.4.
- Deviations: -0.1, 0.1, 0.0, 0.2, -0.2.
- Squared deviations: 0.01, 0.01, 0.00, 0.04, 0.04; sum = 0.10.
- Sample variance = 0.10 / (5-1) = 0.025; sample SD = sqrt(0.025) = 0.158.
- Lab convention: mean to tenths = 12.4; SD to hundredths = 0.16.
- GUM convention: SD to two significant digits = 0.16; mean to same decimal place = 12.40.
Example 2: Data Recorded to Hundredths
Data: 1.23, 1.24, 1.25, 1.26, 1.27.
- Mean = 1.25.
- Sample SD = 0.0158.
- Lab convention: mean 1.25; SD 0.016.
- GUM convention: SD 0.016; mean 1.250.
Counter-Examples
- False precision in the mean: Reporting 12.36 from data measured to 0.1. The raw data cannot support hundredths.
- Lost information in the SD: Reporting 0.2 when the sample SD is 0.158. This hides the spread and may make the result look less precise than it is.
- Mixing conventions in a final uncertainty: 12.4 ± 0.158 is inconsistent. Use 12.4 ± 0.2 or 12.40 ± 0.16.
- Rounding intermediate values: Rounding the variance to 0.03 before taking the square root changes the SD. Keep full precision until final rounding.
Common Mistakes
- Confusing decimal places with significant figures. 12.4 has three significant figures; 0.16 has two significant figures.
- Applying the multiplication and division rule to the mean. The mean is an addition and subtraction result, so use decimal places.
- Treating the standard deviation as an exact number. It is an estimate with its own uncertainty.
- Forgetting the difference between sample and population SD. The rounding rule does not fix a wrong formula.
- Ignoring the reporting context. A classroom lab, a journal article, and a calibration certificate may follow different standards.
- Double rounding. Round only once, at the end of the calculation.
Quick Reference Table
| Situation | Mean | Standard deviation |
|---|---|---|
| Raw data to 0.1, descriptive lab report | 0.1, e.g., 12.4 | 0.01, e.g., 0.16 |
| Raw data to 0.01, descriptive lab report | 0.01, e.g., 1.25 | 0.001, e.g., 0.016 |
| GUM or NIST uncertainty statement | Same decimal place as SD | Two significant digits |
| APA style | Two decimals | Two decimals |
| ASTM E29 or ISO 80000 | Based on significant digits | Based on significant digits |
Practice Problems
- Data: 5.12, 5.15, 5.13, 5.14, 5.16. Compute mean and sample SD. Round for a descriptive lab report with data to hundredths.
- Using the same data, report mean ± SD under the GUM and NIST convention.
- Data: 20, 21, 19, 22, 18. Compute mean and sample SD. Round for a descriptive lab report with data to ones.
Answers: 1. Mean = 5.14; SD = 0.0158, rounded to 0.016. 2. SD = 0.016; mean = 5.140. 3. Mean = 20; SD = 1.58, rounded to 1.6 under the one-extra-decimal convention. Under GUM, SD = 1.6; mean = 20.0.
Related Rules
- Addition and Subtraction: Decimal Places – the rule behind rounding the mean.
- Multiplication and Division: Significant Figures – the rule often misapplied to the SD.
- Standard Deviation Reporting – details on sample versus population SD and reporting conventions.
- Measurement Uncertainty – how SD becomes standard uncertainty in the GUM.
- Rounding Methods – half-up, half-even, and avoiding double rounding.
- Why Sig Figs Are an Approximation – the deeper statistical view.
FAQ
Should the mean and standard deviation always have the same number of decimal places?
No. In descriptive lab reports, the mean often takes the same decimal place as the raw data, while the SD takes one extra decimal place. In GUM and NIST uncertainty statements, the mean and SD should have the same decimal place after the SD is rounded to two significant digits.
Why does my teacher round the mean to one decimal but the SD to two?
Your teacher is likely using the common lab convention: the mean describes the center on the same scale as the measurements, while the SD describes spread and needs an extra decimal place to remain informative. It is a convention, not an arithmetic error.
How many significant figures should a standard deviation have?
For uncertainty reporting, two significant digits are usually sufficient, as stated in GUM 7.2.6 and NIST TN 1297 7.2.6. For descriptive statistics, many fields use two decimal places or one extra decimal place beyond the data.
Does rounding the mean differently change the conclusion?
Rounding alone should not change a scientific conclusion if you keep enough digits in intermediate calculations and round only at the end. Inconsistent rounding, however, can create false precision or hide uncertainty.
What is the safest way to report mean ± SD?
Decide first whether you are making a descriptive statement or an uncertainty statement. For uncertainty, round the SD to two significant digits, then round the mean to the same decimal place. For a lab report, follow your instructor's style guide and state the convention.
Leave a Reply