Definition
The central limit theorem states, under suitable conditions, that a centered and scaled sample mean approaches a normal distribution as sample size grows. For iid observations with finite variance, the scaling is by the population standard deviation divided by the square root of n. The approximation concerns the sampling distribution of the mean, not the raw observations themselves.
How it is calculated
Square root of n times (sample mean minus population mean) divided by population standard deviation converges in distribution to a standard normal.
| Symbol | Meaning | Unit |
|---|---|---|
| n | sample size | count |
| X | observation | measurement units |
| mu | population mean | measurement units |
| sigma | population standard deviation | measurement units |
| N | normal distribution | distribution |
Interpretation
Even when individual returns are not normal, their average can have a nearly bell-shaped sampling distribution when enough suitable observations are combined. The theorem is about repeated samples and scaling. Heavy tails, dependence and small n can make the approximation poor.
Assumptions and conventions
-
The iid finite-variance version assumes independent identically distributed observations with positive finite variance.
-
Other versions require their own dependence and moment conditions.
The standardized statistic is dimensionless; X, mu and sigma retain measurement units.
Limitations and common misreadings
-
The CLT does not say individual returns are normal.
-
There is no universal sample size that makes the approximation adequate.
References
- Larry Wasserman. (2004). All of Statistics: A Concise Course in Statistical Inference. Springer Texts in Statistics. Chapter 5, Convergence of Random Variables. doi:10.1007/978-0-387-21736-9. Source

