Calculate Lognormal Parameters from Mean and Variance
Convert arithmetic mean and variance into the underlying normal distribution parameters for a lognormal model. Instantly compute μ, σ, σ², median, mode, and the coefficient of variation, then visualize the resulting probability density curve.
- Fast conversion from observed moments to log-space parameters
- Built-in validation for positive mean and non-negative variance
- Interactive chart powered by Chart.js for instant interpretation
Lognormal Distribution Graph
This chart plots the probability density function implied by your mean and variance inputs.
How to calculate lognormal parameters from mean and variance
If you need to calculate lognormal parameters from mean and variance, you are usually trying to translate familiar summary statistics into the parameterization used by probability theory. A lognormal distribution describes a positive random variable whose natural logarithm is normally distributed. That simple definition matters because many real-world quantities are nonnegative, right-skewed, and multiplicative in nature. Prices, incomes, claim sizes, particle diameters, rainfall intensities, waiting times, environmental concentrations, and reliability lifetimes often fit a lognormal pattern better than a symmetric normal model.
In practice, analysts often start with the arithmetic mean and variance of the original variable, not with the hidden normal-space parameters. The challenge is that the lognormal distribution is not parameterized directly by those two raw moments. Instead, its standard parameters are the mean of the logarithm, usually written as μ, and the standard deviation of the logarithm, usually written as σ. Once you understand the transformation between the raw scale and the log scale, you can move from observed mean and variance to μ and σ quickly and accurately.
σ² = ln(1 + v / m²)
μ = ln(m) − σ² / 2
Why this conversion matters in applied statistics
The ability to calculate lognormal parameters from mean and variance is valuable in finance, engineering, health sciences, hydrology, and risk analysis. You may know a process average and its spread from reporting standards, simulation outputs, or published research, but the model you need for Monte Carlo simulation or density estimation requires μ and σ. By converting moments into parameters, you can simulate synthetic samples, estimate probabilities of thresholds, construct percentile bands, and compare scenarios on a consistent probabilistic footing.
Another reason this matters is interpretability. The arithmetic mean on the original scale can be heavily influenced by extreme right-tail values, while the median and geometric behavior are often better captured in log space. Converting the parameters helps you move between these viewpoints. It allows you to explain not just the average outcome, but also the skewness, tail behavior, and multiplicative uncertainty in a way that aligns with how many natural and economic systems actually behave.
The formulas behind a lognormal model
Suppose X is lognormally distributed and Y = ln(X) is normally distributed with mean μ and variance σ². Then the raw-scale moments are:
- Mean of X: E[X] = exp(μ + σ² / 2)
- Variance of X: Var(X) = (exp(σ²) − 1) exp(2μ + σ²)
- Median of X: exp(μ)
- Mode of X: exp(μ − σ²)
If your starting values are the arithmetic mean m and variance v, then you solve the system in reverse. First isolate σ² by dividing variance by mean squared:
- v / m² = exp(σ²) − 1
- exp(σ²) = 1 + v / m²
- σ² = ln(1 + v / m²)
Once σ² is known, recover μ using the mean formula:
- m = exp(μ + σ² / 2)
- ln(m) = μ + σ² / 2
- μ = ln(m) − σ² / 2
| Quantity | Symbol | Formula | Interpretation |
|---|---|---|---|
| Arithmetic mean | m | exp(μ + σ² / 2) | The average on the original data scale |
| Arithmetic variance | v | (exp(σ²) − 1)exp(2μ + σ²) | The spread on the original data scale |
| Log-space mean | μ | ln(m) − 0.5 ln(1 + v / m²) | Center of the corresponding normal distribution of ln(X) |
| Log-space variance | σ² | ln(1 + v / m²) | Variance of ln(X), controlling dispersion and skew |
Step-by-step example: calculate μ and σ from mean and variance
Imagine you have a positive-valued process with mean 10 and variance 25. You want the lognormal parameters. Start by computing the ratio v / m². Since m² = 100, you get 25 / 100 = 0.25. Add 1 to obtain 1.25. Now take the natural logarithm:
- σ² = ln(1.25) ≈ 0.2231
- σ = √0.2231 ≈ 0.4723
Next compute μ:
- μ = ln(10) − 0.2231 / 2
- μ ≈ 2.3026 − 0.1116
- μ ≈ 2.1910
From there, you can derive other useful measures:
- Median = exp(μ) ≈ 8.9443
- Mode = exp(μ − σ²) ≈ 7.1572
- Coefficient of variation = √(exp(σ²) − 1) = √0.25 = 0.5
Notice what happened: the mean is larger than the median, and the median is larger than the mode. That is the classic signature of right skew. In a lognormal distribution, the long upper tail pulls the arithmetic mean upward. This is exactly why converting mean and variance to μ and σ is so informative: it reveals the structure underneath the raw moments.
What the coefficient of variation tells you
The coefficient of variation, often abbreviated as CV, is the standard deviation divided by the mean on the raw scale. For a lognormal distribution, it has a particularly elegant expression:
- CV = √(exp(σ²) − 1)
This metric is useful because it normalizes dispersion relative to the level of the variable. In applied fields where variables differ by scale, the CV helps compare relative uncertainty. When the CV is small, the lognormal curve will be relatively concentrated; when it is large, the distribution becomes more heavily skewed with a pronounced upper tail.
Common mistakes when calculating lognormal parameters
One common mistake is using the standard deviation where the formula requires variance. If you are given a standard deviation s, you must square it first to get v = s². Another common error is using base-10 logarithms instead of natural logarithms. The standard lognormal formulas use the natural log, written as ln. Substituting log base 10 will produce incorrect parameter values unless you explicitly convert between bases.
A third mistake is confusing the arithmetic mean on the original scale with the mean of the logarithm. The parameter μ is not the average of the raw data. It is the average of ln(X). These are fundamentally different numbers. For skewed data, the difference can be substantial. Analysts also sometimes forget the validity constraints: the mean must be strictly positive, and the variance must be nonnegative. A lognormal distribution cannot model zero or negative outcomes directly.
| Input situation | Correct action | Avoid this error |
|---|---|---|
| You know mean and standard deviation | Square the standard deviation to get variance before using formulas | Plugging standard deviation directly into the variance formula |
| You use spreadsheet logs | Use natural log functions such as LN() | Using LOG10() or a base-10 calculator function by accident |
| Data include zeros or negatives | Reconsider the model or use a shifted/alternative distribution | Assuming a basic lognormal can fit all numeric data |
| You need simulation inputs | Feed μ and σ into the normal generator in log space | Sampling from a normal distribution directly on the raw scale |
Where lognormal parameter conversion is used
In reliability and survival analysis, lognormal models can represent lifetimes of components where failure mechanisms are multiplicative. In environmental science, concentrations of contaminants and aerosols often exhibit positive skew, making a lognormal structure more realistic than a Gaussian alternative. Public-facing statistical agencies and research institutions often discuss skewed distributions in the context of exposure, health outcomes, and economic measurements. For general statistical guidance and data interpretation resources, you can review materials from the U.S. Census Bureau, the National Institute of Standards and Technology, and introductory probability resources from Penn State University.
In finance, lognormal models are often associated with asset prices because multiplicative returns can produce approximately lognormal price levels under some assumptions. In operations research and supply chain modeling, demand or lead-time variability may be positively skewed. In health economics and insurance, claims or cost distributions can display heavy upper tails, making moment-based conversion essential for scenario analysis and stress testing. Whenever your data are strictly positive and the upper tail matters, the lognormal family becomes a natural candidate.
How to interpret μ and σ after conversion
Once you calculate lognormal parameters from mean and variance, you gain more than just two technical values. The parameter μ locates the center of the distribution in logarithmic space. If you exponentiate μ, you get the median on the raw scale. That makes μ directly tied to a robust central tendency measure. The parameter σ controls the spread in log space, but it also influences skewness, tail thickness, and the separation between the mean, median, and mode on the original scale.
As σ increases, the right tail becomes more pronounced, the arithmetic mean rises relative to the median, and the mode shifts further left. This is why two datasets with the same mean can imply dramatically different risks if their variances differ. The conversion process reveals that distinction clearly. In practical decision-making, this means that averages alone are rarely enough. The same expected value may mask very different distributions of outcomes.
Checklist for accurate calculation
- Confirm the mean is strictly greater than zero.
- Confirm the variance is zero or positive.
- Use natural logarithms, not base-10 logs.
- Compute σ² first with ln(1 + v / m²).
- Take the square root to get σ.
- Compute μ with ln(m) − σ² / 2.
- Optionally derive median, mode, and CV for richer interpretation.
Final takeaway
To calculate lognormal parameters from mean and variance, you do not need iterative estimation or advanced numerical methods. The transformation is exact when the target model is lognormal. Compute σ² = ln(1 + v / m²), then compute μ = ln(m) − σ² / 2. From these, you can derive the median, mode, coefficient of variation, percentiles, and simulation-ready inputs. This conversion is foundational in probability modeling because it bridges descriptive statistics on the original scale with the parameterized distribution needed for serious analysis.
If you work with positive and right-skewed data, this calculation can save time and reduce modeling errors. More importantly, it improves interpretation. Instead of relying only on summary moments, you can understand the hidden log-space structure that drives asymmetry and tail risk. Use the calculator above to turn raw inputs into an actionable lognormal model in seconds.