# Empirical and Mixture Distributions

> A mixture distribution blends several distributions, such as calm and volatile regimes. Learn how mixtures create fat tails, how they are fitted and their uses.

Source: https://learn.tradelabsai.com/math/mixture-distributions/  
Track: Math and Statistics · Level: Advanced · Updated: 2026-10-03  
Publisher: TradeLabs AI (https://tradelabsai.com). Education, not financial advice.  
Cite as: TradeLabs Learn, "Empirical and Mixture Distributions", https://learn.tradelabsai.com/math/mixture-distributions/

Markets do not behave the same way all the time. There are long stretches of calm trading and shorter bursts of turmoil. A mixture distribution models this by combining two or more simpler distributions, each describing a different state, with weights for how often each state occurs. Even if each component is normal, the mixture can have fat tails and skewness, which helps explain why market returns look so different from a single bell curve.

## The idea

A two component normal mixture:

```
f(x) = w × Normal(μ1, σ1) + (1 - w) × Normal(μ2, σ2)
```

- **w:** probability of being in state 1 (for example, calm).
- **1 minus w:** probability of being in state 2 (for example, turbulent).

## Worked example

**Example: Calm and storm**
Suppose daily returns come from a calm regime 90% of the time (mean +0.05%, standard deviation 0.7%) and a stressed regime 10% of the time (mean minus 0.3%, standard deviation 2.5%).

- Overall mean ≈ 0.9 × 0.05% + 0.1 × (minus 0.3%) = +0.015%.
- Overall variance ≈ 0.9 × (0.7² + 0.035²) + 0.1 × (2.5² + 0.315²) ≈ 0.442 + 0.635 ≈ 1.08, so overall standard deviation ≈ 1.04%.

The overall standard deviation is about 1%, but the stressed regime produces 4% and 5% daily losses far more often than a single normal distribution with 1% volatility would. The mixture has fat tails and negative skew, matching real equity behaviour. See [Fat Tails](https://learn.tradelabsai.com/math/fat-tails/).

## Why mixtures produce fat tails

Mixing a narrow distribution with a wide one creates a shape with a tall peak (many quiet days) and heavy tails (occasional extreme days). This is exactly what high kurtosis describes. Volatility clustering, where calm and turbulent periods alternate, can be seen as a dynamic version of this mixing. See [Skewness and Kurtosis](https://learn.tradelabsai.com/math/skewness-and-kurtosis/) and [GARCH](https://learn.tradelabsai.com/math/garch/).

## Fitting mixtures

- **Expectation maximisation (EM) algorithm:** a standard method to estimate the weights, means and standard deviations, based on maximum likelihood. See [Maximum Likelihood](https://learn.tradelabsai.com/math/maximum-likelihood/).
- **Number of components:** chosen using criteria such as AIC or BIC, or by economic reasoning (for example, two regimes: normal and stress).
- **Gaussian mixture models** are also widely used in machine learning for clustering. See [Machine Learning in Trading](https://learn.tradelabsai.com/machine-learning/machine-learning-in-trading/).

## From static mixtures to regime switching

A simple mixture assumes each day independently comes from one state. In reality, regimes persist: a stressed day is usually followed by more stressed days. Regime switching models, such as the Markov switching model introduced to economics by James Hamilton in 1989, let the state evolve over time with transition probabilities. They estimate the probability of being in each regime on each day. See [Structural Breaks and Regime Changes](https://learn.tradelabsai.com/math/regime-changes/).

| Model | Regime persistence | Use |
|---|---|---|
| Static mixture | None | Describing the overall distribution |
| Markov switching | Yes | Detecting and forecasting regimes |
| GARCH | Continuous volatility changes | Volatility forecasting |

## Trading uses

| Use | Example |
|---|---|
| Risk models | More realistic value at risk and stress scenarios. See [Value at Risk (VaR)](https://learn.tradelabsai.com/portfolio/value-at-risk/) |
| Regime filters | Reduce exposure when the model signals a high volatility regime |
| Option pricing | Mixture models can reproduce volatility smiles. See [Volatility Smile and Skew](https://learn.tradelabsai.com/volatility/volatility-smile-and-skew/) |
| Strategy evaluation | Test how strategies perform in each regime |
| Simulation | Generate realistic return paths for Monte Carlo. See [Monte Carlo Simulation](https://learn.tradelabsai.com/research/monte-carlo-simulation/) |

## Limits

- **Regimes are identified after the fact more clearly than in real time.**
- **Parameters can be unstable** with limited data on rare regimes.
- **Too many components** lead to overfitting. See [Overfitting and Curve Fitting](https://learn.tradelabsai.com/research/overfitting-and-curve-fitting/).

## Frequently asked questions

### What is a mixture distribution?

A distribution formed by combining two or more distributions, each representing a different state, weighted by how often each state occurs.

### How do mixture distributions create fat tails?

By combining a narrow distribution for calm periods with a wide one for turbulent periods, producing many quiet values and occasional extreme ones.

### What is a regime switching model?

A model in which the market moves between states, such as calm and volatile, with probabilities that make regimes persist over time.

Next, learn how to analyse data over time in [Time Series Basics](https://learn.tradelabsai.com/math/time-series-basics/).

## Continue learning

- Next lesson: [Time Series Basics](https://learn.tradelabsai.com/math/time-series-basics/)
- Previous lesson: [Skewness and Kurtosis](https://learn.tradelabsai.com/math/skewness-and-kurtosis/)
- Related: [Skewness and Kurtosis](https://learn.tradelabsai.com/math/skewness-and-kurtosis/): Skewness measures whether returns lean to one side; kurtosis measures tail heaviness. Learn the formulas, what they reveal about strategies and how to use them.
- Related: [Structural Breaks and Regime Changes](https://learn.tradelabsai.com/math/regime-changes/): Markets switch between regimes such as calm and turbulent, or trending and ranging. Learn how to detect regimes, the models used and how to adapt strategies.
- Related: [Fat Tails](https://learn.tradelabsai.com/math/fat-tails/): Fat tails mean extreme market moves happen far more often than the normal curve predicts. Learn the evidence, the causes, how to measure them and how to manage them.
- Related: [GARCH](https://learn.tradelabsai.com/math/garch/): GARCH models capture volatility clustering, where big moves follow big moves. Learn the GARCH(1,1) formula, persistence, forecasting and uses in risk and options.
- Related: [Probability Distributions Explained](https://learn.tradelabsai.com/math/probability-distributions/): Probability distributions describe the range and likelihood of outcomes. Learn the main ones used in trading, their shapes and when each applies.
- Related: [Maximum Likelihood](https://learn.tradelabsai.com/math/maximum-likelihood/): Maximum likelihood estimation finds the model parameters that make observed data most probable. Learn the idea, simple examples, its use in GARCH and its limits.