Lesson 7 of 8 · All lessons in this module
- Mean, median, mode and outliers
- Quantiles, box plots and their uses
- Range, mean absolute deviation, variance and standard deviation
- Target downside deviation and coefficient of variationVideo lesson
- Skewness and kurtosisVideo lesson
- Scatter plots, covariance and correlation
- Limits of correlation analysis
- Integrated cases: describing one return series from raw data
Lesson 7 of 8 · 12 min
Limits of correlation analysis
A correlation coefficient is only as good as the data behind it: it misses non-linear links, bends to outliers, can be spurious, and never proves causation.
In short
- Correlation measures only linear association; a strong curved relation can show a correlation near 0.
- Outliers can create a high correlation or hide a real one; decide whether they are noise or information.
- Correlation does not imply causation, however high it is.
- Spurious correlation comes from chance, from dividing both variables by a common third variable, or from both depending on a third variable.
- Identical means, standard deviations and correlations can hide very different data, so always plot the data.
Unlock this lesson free for 7 days
Create a free account to get 7 days free: every lesson, video and flashcard, and the question bank. No card needed.