›Normal Distribution Simulator Back
Statistics

Normal Distribution Simulator

Adjust mean and standard deviation to explore probabilities

Distribution parameters

μ ± 1σ 68.27%
μ ± 2σ 95.45%
μ ± 3σ 99.73%
Results
—
PDF (probability density)
—
CDF (cumulative probability)
—
z-score
—
Deviation score (T-score)
Normal distribution N(μ,σ²)
Theory & Key Formulas
$$f(x) = \frac{1}{\sigma\sqrt{2\pi}} \exp\!\left(-\frac{(x-\mu)^2}{2\sigma^2}\right)$$ $\mu$: mean, $\sigma$: standard deviation, $\sigma^2$: variance

Cumulative distribution function (CDF) and the error function

$$F(x) = \int_{-\infty}^{x} f(t)\,dt = \frac{1}{2}\left[1 + \text{erf}\!\left(\frac{x-\mu}{\sigma\sqrt{2}}\right)\right]$$

z-score and deviation score

z-score: $z = \dfrac{x - \mu}{\sigma}$  Deviation score: $T = 10z + 50$

Basic theory of the normal distribution

The normal distribution through a conversation

🙋
Why do so many kinds of data, like test scores and height, follow a normal distribution?
🎓
There is something called the "central limit theorem". When many independent random factors add up, their sum approaches a normal distribution regardless of the shape of the original distributions. Height is the accumulation of thousands of factors such as genes, nutrition and environment. Test scores also depend on many factors, such as knowledge, physical condition and how well you get on with the questions. So they naturally become bell-shaped.
🙋
I often hear about "3-sigma control" in factories. Is that related to the normal distribution?
🎓
That is exactly an application of the 68-95-99.7 rule of the normal distribution. For a normal distribution the probability of falling outside ±3σ is 0.27% (about 27 in 10,000). If quality control accepts "within 3σ", the defect rate is 0.27%. Six Sigma is a quality target that tightens this to ±6σ (3.4 per million) and is widely used in automotive and semiconductor manufacturing.
🙋
I understand that a deviation score of 50 is "average", but how rare is a deviation score of 70?
🎓
A deviation score of 70 corresponds to z = 2.0. For a normal distribution P(z > 2.0) ≈ 2.3%, so it means being in roughly the top 2.3% of all test-takers. A score of 80 is z = 3.0, with P(z > 3.0) ≈ 0.13%, about 1.3 people in 1000. Set μ = 0 and σ = 1 in this tool and put the evaluation point at 2 or 3, and the CDF shows how rare that is.
🙋
How is the normal distribution used in CAE? Is it separate from the finite element method?
🎓
They are closely related. For example, in structural reliability analysis the probability of failure Pf = P(Z < 0) is calculated from the distribution of the difference Z = R − σ between the stress σ (normally distributed) and the material strength R (normally distributed). In Monte Carlo simulation, normal distributions are assigned to the input variables, many analyses are run and the distribution of the outputs (stress, displacement and so on) is estimated. The normal distribution is also the basis of dimensional tolerance analysis (tolerance stack-up) of products.

Frequently Asked Questions

In many cases yes: by the central limit theorem, the distribution of the sample mean approaches a normal distribution when the sample size is 30 or more. With severe skew or outliers, however, non-parametric tests (such as the Mann–Whitney test) are more appropriate.
The log-normal distribution is used when ln(X) follows a normal distribution. It suits data that "cannot be zero or negative and have a heavy tail", such as material strength, fatigue life, earthquake magnitude and financial prices. In FEM analyses of material scatter, fatigue strength is often assumed to be log-normal.
There is no exact closed form; numerical integration or the error function erf(x) is used. A common approximation is the rational approximation of Abramowitz & Stegun (error < 1.5×10⁻⁷). This tool uses a rational approximation to calculate the CDF with high accuracy.
The standard normal distribution is the special normal distribution standardised to μ = 0 and σ = 1. Any normal distribution N(μ, σ²) can be converted to it with z = (x − μ)/σ. With a z-table (standard normal table), the probabilities of any normal distribution can be computed, which is why it was widely used in the days of hand calculation.

What Is the Normal Distribution and Statistics Calculator?

The normal distribution is described by a bell-shaped probability density function (PDF), symmetric about the mean μ, \( f(x) = \frac{1}{\sqrt{2\pi\sigma^2}} e^{-\frac{(x-\mu)^2}{2\sigma^2}} \). In this simulator you control μ and the standard deviation σ with sliders and see the change in the shape of the distribution immediately. For example, a smaller σ makes the curve sharper and a larger σ flattens it. The cumulative distribution function (CDF) is defined as \( F(a) = \int_{-\infty}^{a} f(x) dx \) and gives the probability of a given interval. Following the 68-95-99.7 rule, you can check visually that about 68% of the data lie within μ±σ, about 95% within μ±2σ and about 99.7% within μ±3σ. The z-score standardises a value as \( z = \frac{x-\mu}{\sigma} \), making it easy to compare different normal distributions. The deviation score is the value converted to a mean of 50 and a standard deviation of 10, calculated from the z-score as \( T = 50 + 10z \). Observing the probability density and cumulative probability in real time while changing these parameters gives an intuitive grasp of the basics of statistical inference.

$f(x) = \frac{1}{\sigma\sqrt{2\pi}} \exp\!\left(-\frac{(x-\mu)^2}{2\sigma^2}\right)$

Real-World Applications

Industrial use (automotive)
On automotive component production lines, dimensions such as the outer diameter of piston rings are modelled with a normal distribution, the process capability indices (Cp/Cpk) are monitored, and the probability of defects exceeding the upper specification limit is estimated from the z-score. It also helps detect early a shift of the mean caused by tool wear and gives a basis for deciding when to replace tools.

Research and education
University statistics courses often include the exercise "what percentage of students have a deviation score of 60 or more?", treated as a normal distribution with μ = 50 and σ = 10 and checked from the CDF value of 84.13% for z = 1.0 (top 15.87%). In this tool you can confirm the same value by entering μ = 0, σ = 1 and x = 1. In pharmacy it is also used to model the distribution of blood concentrations in clinical trial data and estimate the probability of falling within the effective concentration range.

Relation to CAE analysis and practical use
In CAE such as crash analysis, material properties like the yield strength of steel sheet are treated as normally distributed (for example μ = 350 MPa, σ = 15 MPa), and the design margin is sometimes checked against a lower value such as μ − 3σ (305 MPa). This is the basis of robust design that takes scatter into account.

Common Misconceptions and Points to Note

It is true that "the larger the standard deviation, the larger the scatter of the data", but it is tempting to think that "for normally distributed data, about 68% of the data always lie within the mean ±1σ". In fact this proportion is only a theoretical value, and when the sample size is small or the data deviate from a perfect normal distribution, the measured proportion can depart considerably from 68%. In practice the tails of the distribution are often extended by outliers and measurement errors, so do not over-trust the 68-95-99.7 rule; interpret it while checking the actual distribution of the data.

Also, the understanding that "a deviation score of 50 is the mean and 60 is about the top 16%" is correct, but it is tempting to think that "the higher the deviation score, the better". In reality the deviation score only shows the relative position within that group and cannot simply be compared between different tests. For example, the same deviation score of 60 on two tests taken by groups of different levels can correspond to completely different levels of ability.

How to Use

  1. Set "Mean μ" in the range −5 to 5. Real dimensions can be handled by converting them to z-scores (standardised values) before entering them (for example, to evaluate 101 mm for a part with mean 100 mm and σ 0.5 mm, enter μ = 0, σ = 1, x = 2)
  2. Set "Standard deviation σ" in the range 0.2 to 4. If quality control uses ±3σ as the control range, σ = 0.5 mm gives a control range of 100 ± 1.5 mm
  3. Enter the value to evaluate as "Evaluation point x" and its z-score (standardised value), probability density PDF, cumulative probability CDF and deviation score are calculated in real time

Worked Example

For automotive engine parts made with an outer diameter of mean μ = 50.00 mm and standard deviation σ = 0.08 mm, a measured value x = 50.10 mm has a z-score of (50.10 − 50.00)/0.08 = 1.25 and a CDF value of about 0.894 (lower 89.4%). For x = 49.90 mm, z = −1.25 and the CDF value is about 0.106. Because the tool's μ and σ ranges (μ: −5 to 5, σ: 0.2 to 4) do not allow the actual dimensions to be entered directly, enter μ = 0, σ = 1 and evaluation point x = 1.25 to get the same z-score and CDF (89.44%). By the 68-95-99.7 rule, 68.3% of all parts lie within ±1σ (49.92–50.08 mm).

Practical Notes

  1. In pharmaceutical content control, manufacturing conditions are adjusted so that 99.73% (the 3σ range) fall within the specification of ±10% of the stated content, and the process is evaluated with the process capability Cpk = (USL − μ)/3σ (in this tool, setting the upper specification limit USL as the evaluation point x shows from the CDF the proportion exceeding the specification)
  2. For weight control in food production with a mean of 500 g and σ = 2 g, the defect rate below 494 g (z = −3) is about 0.13% and below 498 g (z = −1) about 15.9%. In this tool you can confirm this by entering μ = 0, σ = 1 and x = −3 or −1
  3. If the σ entered is too small the confidence interval becomes unjustifiably narrow, so check normality beforehand with a goodness-of-fit test for the normal distribution (Shapiro–Wilk test) on past data