Hi Eetu
EP: You are still not answering but just throw the question back to me.
RM: Sorry, Eetu. I was just asking what you thought were Adam’s “strong arguments that the empirical dependence between curvature and velocity cannot generally (and logically) be a mathematical, statistical or measurement created artefact”. I can’t explain why I think those arguments are wrong unless you tell me what they are.
EP: Perhaps you have no input functions for criticism and counter arguments? That would explain why you ask me to repeat them again.
RM: Why all the anger? I didn’t ask you to repeat your question. You never told me what Adam’s arguments were that convinced you that the empirical dependence between curvature and velocity cannot be a mathematical, statistical or measurement created artefact. I just asked you what those arguments were.
EP: But OK will repeat the most important argument with my own words: Even a child knows that if you have a simple normal coordinate space and a moving point in it then the velocity of the moving point can no way depend on the place of the point or on the direction of its movement. They are independent phenomena. Adam just showed this with data and diagrams. If there is a (real) change in the velocity of the point it is not caused by mathematics or statistics. You have a real phenomenon which requires a real explanation.
RM: I know that the power law is a real phenomenon. And what I have shown is that velocity is mathematically dependent on curvature, not direction. This is just a mathematical fact, acknowledged even by power law researcher themselves (1,2) The relationship is described by this equation:
(1)
where A is angular velocity, C is curvature and D is something called affine velocity.
RM: The power law is the empirical finding of a 2/3 power relationship between C (curvature) and A (velocity) in the movements produced by living organisms. That is, the power law is the finding that the empirical relationship between measurements of C and A is:
(2)
where beta is typically close to 2/3.
RM: But note that there is a 2/3 power relationship between C and A in the mathematical relationship between C and A (equation 1)! The only difference between equation 1 – the mathematical relationship between C and A – and equation 2 – with beta = 2/3 – is the variable D in equation 1. Apparently this remarkable “coincidence” – the similarity between equations 1 and 2 – didn’t make a big impression on power law researchers but it made a big impression on me.
RM: I knew that the empirical power law – equation 2 – was found using simple linear regression, where the log of the observed values of C were regressed on the logs of the observed values of A to determine the best fitting power law relationship between curvature and velocity. That is they do a regression analysis using the equation:
(3)
to find the value of beta that gives the best fit to a power relationship between observed values of C and A. The regression analysis typically results in a value of beta nearly equal to 2/3 and an R^2 value greater than .9 (indicating a very good fit to the power equation.
RM: But it is clear from equation 1 that the results of this regression are artifactual because the regression analysis leaves out one of the variables on which the value of A depends; it leaves out the variable D, the measure of affine velocity. The appropriate regression analysis for predicting log (A) from log (C) would be a multiple regression analysis that include two predictors of log(A): log (C) and log (D). When you do this analysis what you always find is that the beta for predicting log (A) from log (C) is exactly 2/3, because that’s what equation 1 says it is mathematically. That is, you always find that the regression equation for predicting log(A) from log(C) is exactly what you would expect from equation 1:
log (A) = 1∕3 ⋅ log (D) + 2∕3 ⋅ log (C ) (4)
This is true for all movement trajectories, even those of the planets in their orbits.
RM: What this means is that the empirical power “law” – the apparent decrease in velocity as the 2/3 power of curvature – depends on variations in a variable – affine velocity – that is not included in the typical regression analysis that is used to find the power “law”. If affine velocity is nearly constant – as it apparently is in most trajectories (2) – then the exponent of the power relationship found by regressing C on A (omitting D) will be close to the true mathematical value of this exponent (per equation 1).
RM: Another way of saying it is like this: The 2/3 power law decrease in movement velocity (A) as a function of movement curvature (C) is a mathematical fact (per equation 1). The degree to which this mathematical relationship is observed in actual movement, by regressing measures of C on measures of A, depends on the degree to which the affine velocity of the movement (the variable D) is constant and, therefore, uncorrelated with variation in curvature, C.
RM: I’m not aware of anyone – let alone Adam – showing that there is anything wrong with this analysis. But maybe I missed it so feel free to show me where I am mistaken.
EP: As for your question about the relevance of the power law to PCT, I must say I don’t know. Power law phenomenon is an interesting phenomenon for research by its own right. I think it is great the Adam tries to apply PCT to that research.
PCT is relevant for it like for many other questions but it is just a dream that PCT could just annihilate those questions away from science.
RM: PCT doesn’t annihilate questions away from science (whatever that means). PCT explains the power law as an irrelevant side effect of movement which can’t tell you anything about how that movement was produced. So I guess PCT does kind of annihilate the power law as a phenomenon that can tell you anything about how movements are produced; but PCT does show that the power law is an interesting example of a behavioral illusion that has kept researchers going down a blind alley for many years.
References
1.Pollick F, Sapiro G (1997) Constant affine velocity predicts the 1/3 power law of planar motion perception generation. Vision Res, 37:347–353
2. Maoz U, Portugaly E, Flash T, Weiss Y (2006) Noise and the 2/3 power
law. Adv Neural Inf Proc Syst 18:851–858