patterns in randomness: 2

An example of seeing patterns relates to the way we tend to segment data. If we are presented with the following data:

we may segment the data like this:

and infer that as x increases, y increases (i.e. x is correlated to y).

We must remember however that:

Correlation does not imply causation

Our segmentation of the data is totally arbitrary. If we split the data like this:

we would see no pattern at all.

This segmentation effect is common in stock markets where a falling stock with often hover around a psychologically significant number (1000, 5000 etc.) for some time before ‘breaking through’ or ‘recovering’. We must remember that these numbers are arbitrary segmentation points. If human psychology was not involved then there is no reason to assume any magical significance for 2000 rather than 2347.6

The way we naturally segment data can lead to incorrect decisions or to not making decisions when a different segmentation point would make the decision clear. Decision makers should carefully consider whether their segmentation points are arbitrary or whether other points would prove more useful.

Aricle link