An example of how to decide if 2 time series are correlated or not.

Graphically speaking, one can estimate the correlation of 2 time series by using a scatter plot of those 2 time series (right top graph): the more the points are aligned in a straight line, the more the time series are correlated.

Computationnaly speaking, the **coefficient of correlation** gives an insight of the dispersion of those points: the more the coefficient of correlation is near 1 (or -1), the more the scatter points are aligned, and the more the 2 time series are correlated.

- in the left graph, Drag & Drop a point to see the impact on the coefficient of correlation
- in the left graph, Drag & Drop a timeline (to change each values of the related time serie), and see that it has
**no impact**on the coefficient of correlation; - in the left graph, correlate or decorrelate time serie 2 and time serie 1 by introducing some random values; then see the impact on the coefficient of correlation;
- in the left graph, increase or decrease the trend of the second time serie, and see that it has
**no impact**on the coefficient of correlation; - in the left graph, inverse a trend, and see that it inverses the coefficient of correlation; when the 2 trends increases/decreases in the same way, the coefficient of correlation is positive; at the opposite, when one trend is increasing and the other is decreasing, the coefficient of correlation is negative;
- in the left graph, disperse/concentrate the second time serie, and see that the more disperse is the second timeserie, the less it correlates with the first time serie;

- trend line computed using
*least square method*

- done with D3 v3.5.5
- blockbuilder.org