01
Examine the relationship
First, check how closely the two time series move together.
Does that apparent lead really hold up?
CauseDecode is an analysis tool for testing relationships between two time series.
It looks beyond whether they move together to ask which one tends to move first.
By combining multiple forms of evidence, it shows how strongly the hypothesis is supported by the data.
Moving together isn't the whole story.
-
They're correlated. Does that mean the relationship is real?
- Even when two series move together, the pattern may be coincidental or driven by another factor affecting both.
-
A moves first. Is it really leading?
- Change the sample period or lag settings, and an apparent lead can disappear—or even reverse direction.
-
Does the relationship hold up over time?
- A relationship that appears only in one period should not be treated the same as one that is observed repeatedly across different windows.
How we test
One number doesn't define the relationship.
It separates relationship, timing, predictive direction, and stability,
then organizes the evidence behind the hypothesis.
02
Examine the timing
Shift one series forward and backward to see where the relationship appears strongest.
03
Test predictive direction
Test whether past values of A add information when predicting B, separately from the reverse direction.
04
Check stability
Check whether the relationship is limited to one period, then organize the evidence as a whole.
Look beyond relationship strength to timing, direction, and stability.
Co-movement
Lead-lag
Predictive direction
Stability