Computing Pearson's correlation by knowing the differences between consecutive values

57 Views Asked by At

Let's say we have two variables X and Y with n observations for each variable. We don't know the values of each observation, but we are given the difference between consecutive observation of each variable. For example:

X=[a,b,c], Y=[d,e,f] (a,b,c,d,e,f are unknown)

[b-a,c-b] and [e-d,f-e] are given and we know the values of b-a, c-b, e-d, f-e.

Is there any way to compute Pearson's correlation coefficient among X and Y by using this information? (one simple way could assign a constant value to a and d and construct observations by adding the given differences to each value and computing the Pearson correlation, but I'm looking for another approach which only uses the differences)

1

There are 1 best solutions below

0
On

The Pearson correlation is invariant for translations, so the correlation of X and Y as defined in your example is the same as the correlation of [0,b-a,c-a] and [0,e-d,f-d]. And note that these last two can be written as [0,b-a,(c-b)+(b-a)] and [0,e-d,(f-e)+(e-d)]. So you can compute everything in terms of the differences.