I am a first-year undergrad student and I have been reading about Transfer Entropy for my research. Although I understand the math behind I am not really sure what the value means. For example, I run $TE(X\mapsto Y)$ and I get the value of TE as 0.624.
What does this value mean?
$\begingroup$
$\endgroup$
0
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
Since you've already followed the math, I'll provide an explanation in intuitive, layman terms rather than the precise definition.
$TE(X\mapsto Y)=0.624$ means that the history of the X process has 0.624 bits of additional information for predicting the next value of Y. (i.e., it provides information about the future of Y, in addition to what we know from the history of Y).
Since it is non-zero, you can conclude that X influences Y in some way.
The TE should not be negative. To see if the TE is actually significant, you could compare it with the entropy of Y. If H(Y) is, say 200 bits per sample, 0.624 may not be significant. TE should never be more than the entropy of Y, so that's the range you expect.
answered Jul 24, 2018 at 15:45
-
$\begingroup$ I have two questions: 1) can TE be negative? 2) Is there a range in which TE oscillates? $\endgroup$ Commented Jul 24, 2018 at 17:58
-
$\begingroup$ Just added explanation in the answer. $\endgroup$ Commented Jul 24, 2018 at 22:49
-
$\begingroup$ could you explain how in the above case TE is more than the Entropy of Y? Is there a general ratio we assume when comparing TE with the corresponding entropy? $\endgroup$ Commented Jul 27, 2018 at 20:40
-
$\begingroup$ TE should not be more than the entropy. It's either a machine floating point error or a bug. There isn't a general rule for the ratio that I know of - it's something you might consider as it relates to the problem. $\endgroup$ Commented Jul 30, 2018 at 19:50