Knut Larsen: Demystifying Correlation

1
Knut LarsenDemystifying Correlation

• Simple techniques for diagnosing and quantifying
  correlation

2
Why Diagnose Correlation

• Selecting investments for portfolio
  diversification
• Selecting investments for appropriate sector
  representation

3
Why Quantify Correlation

• Determine the strength of correlation
• Rank investments based on correlation strength
• Select most appropriate investments

4
The Raw Data
5
Time-Series Charts

• Show the data in sequence
• Easy to understand and interpret
• May indicate type of correlation

6
Kelver versus VennesTime-series Chart
7
Scatter Diagrams

• Readily show type of correlation
• Indicate strength of correlation
• Easily accommodate data at different levels

8
Kelver versus VennesScatter Diagram
9
Positive Correlation

• Data are plotted around a line that is sloped
  upward to the right (positive slope)
• The closer to the line the data plots are, the
  stronger is the correlation

10
Kelver versus GosvinTime-Series Chart
11
Kelver versus GosvinScatter Diagram
12
Negative Correlation

• Data are plotted around a line that is sloped
  downward to the right (negative slope)
• The closer the data plots are to the line, the
  stronger is the correlation

13
Maglin versus SelforTime-Series Chart
14
Maglin versus SelforScatter Diagram
15
Maglin versus SelforScatter Diagram
16
Uncorrelated data

• Data plots are widely spread
• We may draw a line that may be sloped upward or
  downward to the right, or with no slope at all
• Since data are so spread out, the correlation is
  weak

17
Tallem versus SelforTime-Series Chart
18
Tallem versus SelforScatter Diagram
19
Correlation coefficient
20
Positive Correlation
I
III
II
42.67
21
Negative Correlation
40.58
42.67
22
No (or low) Correlation
40.58
42.75
23
Quantifying Correlation

• If there is perfect, positive correlation, the
  correlation coefficients value is 1
• If there is perfect, negative correlation, the
  correlation coefficients value is 1
• If there is no correlation, the correlation
  coefficients value is 0

24
Correlation 6 Stocks
25
Correlation 6 Stocks
26
Required Correlation
When is there enough correlation to say that
there is a linear relationship?
General Rule of Thumb - a linear relationship
exists if
i.e. no linear relationship between Tallem and
Selfor, since rxy 0.5265 lt
27
Correlation in Excel
28
Correlation in Excel
29
Correlation in Excel
click
30
Correlation in Excel
click
31
Correlation in Excel
click
32
Correlation in Excel
33
Correlation in Excel
Enter
34
Correlation in Excel
click
35
Correlation in Excel
the correlation coefficient for Kelver and Vennes
36
The End Thank you for your attention