Statistics describes a set of returns with a few numbers: where the centre is, how spread out the values are, and what shape they make.
Measures of central tendency
- Arithmetic mean: the sum divided by the count.
- Median: the middle value when sorted (the average of the two middle values if the count is even). It is not pulled around by extreme values.
- Mode: the most frequent value. A distribution can have more than one.
- Trimmed and winsorized means handle outliers: a trimmed mean drops a percentage of the highest and lowest values; a winsorized mean replaces them with the nearest remaining values.
Quantiles split sorted data into equal parts: quartiles (4), quintiles (5), deciles (10), percentiles (100). The position of the y-th percentile in n sorted values is approximately (n + 1) × y / 100.
Measures of dispersion
Take returns of 4%, 8%, −2% and 10%. The mean is 5%.
| Measure | Formula | Example |
|---|---|---|
| Range | Highest − lowest | 10 − (−2) = 12% |
| Mean absolute deviation | Average of absolute deviations from the mean | (1 + 3 + 7 + 5) / 4 = 4.0% |
| Sample variance | Σ(x − mean)² / (n − 1) | 84 / 3 = 28 |
| Sample standard deviation | √variance | √28 = 5.29% |
Dividing by n − 1 for a sample (not n) corrects for the fact that the sample mean is itself estimated from the data. The calculator’s statistics worksheet gives both: Sx (sample) and σx (population).
Target downside deviation measures only the bad side: it uses deviations below a target return (often zero or a minimum acceptable return) and ignores those above it.
Coefficient of variation (CV) = standard deviation / mean. It shows risk per unit of return. Here CV = 5.29 / 5 = 1.06. When comparing two assets, the lower CV offers less risk for each unit of return.
Shape: skewness
Skewness measures asymmetry.
- Symmetrical (like the normal distribution): mean = median = mode.
- Positively skewed: a long tail to the right. A few big gains pull the mean up, so mean > median > mode. Many individual stocks look like this.
- Negatively skewed: a long tail to the left, so mean < median < mode. Strategies that sell options often look like this: many small gains, occasional big losses.
Investors generally prefer positive skew, because the surprises are on the upside.
Shape: kurtosis
Kurtosis measures how much of a distribution sits in its tails compared with a normal distribution. A normal distribution has a kurtosis of 3, so analysts usually quote excess kurtosis = kurtosis − 3.
- Leptokurtic (excess kurtosis > 0): fatter tails. Extreme returns happen more often than a normal model predicts. Most asset returns are like this.
- Mesokurtic (= 0): normal.
- Platykurtic (< 0): thinner tails.
The practical lesson: models that assume normal returns underestimate the chance of crashes.
Correlation, briefly
Covariance measures whether two variables move together; its size depends on units, so it is hard to read. Correlation standardizes it to a range of −1 to +1: correlation = covariance / (σx × σy). Correlation measures only linear relationships, and it can be distorted by outliers. It also does not imply causation. Correlation reappears in portfolio maths in lesson 5.
Exam focus
Expect to calculate a mean, a sample standard deviation or a CV from a few numbers, to rank assets by CV, and to identify skew from the order of mean, median and mode. Remember: sample variance divides by n − 1.