NuclearPhynance Archive

I can only repeat what I said in the Virtual Reality thread. While all this stuff looks nice, you can apply a transformation to any n-Dimensional dataset to reduce it to a vector or further into a scalar, which will have to be done anyway for trading. In trading, less is more and I can guarantee you that nothing you can spot in 5 dimensions can't be spotted easier in 2 dimensions after the proper transformations are applied.

i have been thinking about what you wrote. i agree with your statement that less is more for trading. i either don't understand what you are trying to say or i simply disagree with the rest.



i don't think that a data set can be transformed from n-D to 2-d (n>2) without losing information. sometimes that is either fine or even better. but in certain situations, adding the dimensions reduces the complexity of the problem. consider the person from "flatland" (although i have not read the book, i am familiar with its point). consider how hard it would be to identify a 3-d cube while only looking at a 2-d transformation. and how easy it is in 3-d. truly, this is an extreme example. but i am sure that there are plenty of examples where one variable taken alone has little explanatory value but combined with another has much.



my guess is that your answer will be to transform the two into one variable and use that. fine. but how was that relationship determined? and that is my point. why not put another tool in the box? what if a multi-dim visualization and some hours using it can bring a little bit of enlightenment toward creating the next transform? my personal experience with par coord has been quite positive. i have found the interaction with the data conducive to the way i learn about data. perhaps others do not share my learning mode.



maybe i am missing something regarding how you determine the transformations.