Geometric Standard Deviation Calculator

A geometric standard deviation calculator works out how spread out a set of positive numbers is around their geometric mean, using the natural logarithm of each value. You enter a list of numbers (all greater than zero) and the calculator returns the geometric standard deviation as a dimensionless multiplicative factor, along with the geometric mean and the number of entries. This is the correct measure of dispersion for data that grows or decays exponentially, such as investment returns, dilution series, or population growth over time.

Formula

The geometric standard deviation (GSD) is defined as the exponential of the ordinary sample standard deviation of the natural logarithms of the data. For a list of positive values \(x_1, x_2, \ldots, x_n\):

\(\text{GSD} = \exp\left( \sqrt{ \frac{1}{n-1} \sum_{i=1}^n \left( \ln(x_i) – \overline{\ln x} \right)^2 } \right)\)

where \(\overline{\ln x}\) is the arithmetic mean of the natural logs of the data. The geometric mean is \(\exp(\overline{\ln x})\). All values must be strictly positive because the logarithm of zero or a negative number is undefined. The result is a dimensionless factor (no unit) that is always greater than or equal to 1.

Worked Example

Using the default values 1.5, 2.3, 3.1, 4.2, 5.0:

1. Take natural logs: ln(1.5)=0.4055, ln(2.3)=0.8329, ln(3.1)=1.1314, ln(4.2)=1.4351, ln(5.0)=1.6094.
2. Mean of logs = (0.4055+0.8329+1.1314+1.4351+1.6094)/5 = 1.0829.
3. Squared deviations: (0.4055-1.0829)²=0.4588, (0.8329-1.0829)²=0.0625, (1.1314-1.0829)²=0.0024, (1.4351-1.0829)²=0.1240, (1.6094-1.0829)²=0.2772. Sum = 0.9249.
4. Sample variance = 0.9249/(5-1) = 0.2312.
5. Sample standard deviation of logs = √0.2312 = 0.4808.
6. GSD = exp(0.4808) = 1.617 (rounded to 3 decimals).

The geometric mean is exp(1.0829) = 2.953. So the data is roughly centered at 2.95, and one GSD above is 2.95 × 1.617 = 4.77, one GSD below is 2.95 / 1.617 = 1.83.

When to Use the Geometric Standard Deviation

Use the GSD instead of the ordinary (arithmetic) standard deviation when your data is multiplicative or log-normal. Typical cases:

• Investment returns over multiple periods (e.g., annual returns of 10%, −5%, 15%).
• Concentrations in serial dilutions (e.g., 1:2, 1:4, 1:8).
• Growth rates of populations, bacteria, or cells.
• Ratios or indices where values are always positive and span several orders of magnitude.

If your data can be negative or zero, the GSD is not defined. In that case use the arithmetic standard deviation or transform the data first (e.g., add a constant).

How to Read the Result

The GSD is a multiplicative factor, not an additive amount. If the geometric mean is \(m\) and the GSD is \(g\), then approximately 68% of the data falls between \(m/g\) and \(m \times g\), and about 95% falls between \(m/g^2\) and \(m \times g^2\) (for log-normal data). For example, a GSD of 1.5 with a geometric mean of 10 means most values lie between 6.7 and 15, not between 8.5 and 11.5 as an arithmetic standard deviation would suggest. Always compare GSDs relative to each other: a GSD of 2.0 is twice as spread out as a GSD of 1.5 in multiplicative terms.

Common Mistakes and Misconceptions

• Using GSD with negative or zero values: The calculator returns an error or undefined result. Filter your data first.
• Confusing sample vs population: The calculator uses the sample standard deviation (dividing by n−1) by default, which is correct when your data is a sample from a larger population. If your list is the entire population, use the population GSD (divide by n).
• Interpreting GSD as an additive spread: A GSD of 1.2 does not mean “plus or minus 0.2â€� in the original units; it means “multiply or divide by 1.2â€�.
• Rounding the logs too early: Keep at least 4 decimal places during intermediate steps to avoid a noticeably wrong final answer.
• Using the arithmetic mean in the formula: The GSD is based on the mean of the logs, not the mean of the original values.

Limitations of the Geometric Standard Deviation

The GSD is only meaningful for positive data that is approximately log-normal. If your data has a different distribution (e.g., uniform or normal), the GSD may not be a useful summary. Also, the GSD is highly sensitive to extreme values when the sample size is small—with fewer than 5 data points, the result can be unstable. For very large datasets, the GSD tends to be more reliable. Finally, the GSD does not tell you anything about skewness or the shape of the distribution beyond spread; always plot your data alongside the numeric summary.

Alternative Measures of Spread

Depending on your data, one of these may be more appropriate:

• Arithmetic standard deviation: For data that is symmetric and can be negative.
• Median absolute deviation (MAD): Robust to outliers, works for any distribution.
• Interquartile range (IQR): The range between the 25th and 75th percentiles, good for skewed data.
• Coefficient of variation (CV): The arithmetic standard deviation divided by the mean, useful for comparing relative spread across different scales.

Frequently Asked Questions

What is the difference between geometric standard deviation and regular standard deviation?

The regular standard deviation measures spread in additive terms (e.g., ±5 units), while the geometric standard deviation measures spread in multiplicative terms (e.g., ×1.5 or ÷1.5). The GSD is the exponential of the standard deviation of the log-transformed data, so it only works for positive numbers.

Can I use the geometric standard deviation for negative numbers?

No, because the logarithm of a negative number is undefined. If your data contains negatives, you must either shift the data by adding a constant to make all values positive, or use a different measure of spread like the arithmetic standard deviation.

How do I calculate the geometric standard deviation in Excel?

In Excel, you can use the formula =EXP(STDEV.S(LN(range))) for a sample, or =EXP(STDEV.P(LN(range))) for a population. Enter the data in a column, then apply the LN function to each cell, then take the standard deviation of those logs, then exponentiate.

What does a geometric standard deviation of 1 mean?

A GSD of exactly 1 means there is no spread at all—all values in the data set are identical. In practice, a GSD close to 1 (e.g., 1.05) indicates very little variability, while a GSD of 2 or more indicates substantial multiplicative spread.

Should I use sample or population geometric standard deviation?

Use the sample version (n−1 denominator) when your data is a random sample from a larger population, which is the most common scenario in statistics. Use the population version (n denominator) only when your data covers the entire population, such as all students in a specific class or all products from a single production batch.

Why is the geometric standard deviation always greater than or equal to 1?

Because it is the exponential of a standard deviation, which is always non-negative. The exponential of 0 is 1, and the exponential of any positive number is greater than 1. Thus the GSD is always ≥1, with 1 meaning no variability.

Related

• Expected Value Standard Deviation Calculator
• Pace Calculator: Min per Km to Speed
• Exam Average Calculator: How to Find It

in list | Values (comma separated) | 1.5, 2.3, 3.1, 4.2, 5.0 | 0
out gsd | Geometric standard deviation | - | exp(sd(list)) | 3
out gmean | Geometric mean | - | exp(mean(list)) | 3
out n | Count | - | count(list) | 0