02.07.2014 Views

Lecture Notes on Compositional Data Analysis - Sedimentology ...

Lecture Notes on Compositional Data Analysis - Sedimentology ...

Lecture Notes on Compositional Data Analysis - Sedimentology ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

12 Chapter 3. Aitchis<strong>on</strong> geometry<br />

two operati<strong>on</strong>s which give the simplex a vector space structure. The first <strong>on</strong>e is the<br />

perturbati<strong>on</strong> operati<strong>on</strong>, which is analogous to additi<strong>on</strong> in real space, the sec<strong>on</strong>d <strong>on</strong>e is<br />

the power transformati<strong>on</strong>, which is analogous to multiplicati<strong>on</strong> by a scalar in real space.<br />

Both require in their definiti<strong>on</strong> the closure operati<strong>on</strong>; recall that closure is nothing else<br />

but the projecti<strong>on</strong> of a vector with positive comp<strong>on</strong>ents <strong>on</strong>to the simplex. Sec<strong>on</strong>d, we<br />

can obtain a linear vector space structure, and thus a geometry, <strong>on</strong> the simplex. We just<br />

add an inner product, a norm and a distance to the previous definiti<strong>on</strong>s. With the inner<br />

product we can project compositi<strong>on</strong>s <strong>on</strong>to particular directi<strong>on</strong>s, check for orthog<strong>on</strong>ality<br />

and determine angles between compositi<strong>on</strong>al vectors; with the norm we can compute<br />

the length of a compositi<strong>on</strong>; the possibilities of a distance should be clear. With all<br />

together we can operate in the simplex in the same way as we operate in real space.<br />

3.2 Vector space structure<br />

The basic operati<strong>on</strong>s required for a vector space structure of the simplex follow. They<br />

use the closure operati<strong>on</strong> given in Definiti<strong>on</strong> 2.3.<br />

Definiti<strong>on</strong> 3.1. Perturbati<strong>on</strong> of a compositi<strong>on</strong> x ∈ S D by a compositi<strong>on</strong> y ∈ S D ,<br />

x ⊕ y = C [x 1 y 1 , x 2 y 2 , . . .,x D y D ] .<br />

Definiti<strong>on</strong> 3.2. Power transformati<strong>on</strong> of a compositi<strong>on</strong> x ∈ S D by a c<strong>on</strong>stant α ∈ R,<br />

α ⊙ x = C [x α 1 , xα 2 , . . .,xα D ].<br />

For an illustrati<strong>on</strong> of the effect of perturbati<strong>on</strong> and power transformati<strong>on</strong> <strong>on</strong> a set<br />

of compositi<strong>on</strong>s, see Figure 3.1.<br />

The simplex , (S D , ⊕, ⊙), with the perturbati<strong>on</strong> operati<strong>on</strong> and the power transformati<strong>on</strong>,<br />

is a vector space. This means the following properties hold, making them<br />

analogous to translati<strong>on</strong> and scalar multiplicati<strong>on</strong>:<br />

Property 3.1. (S D , ⊕) has a commutative group structure; i.e., for x, y, z ∈ S D it<br />

holds<br />

1. commutative property: x ⊕ y = y ⊕ x;<br />

2. associative property: (x ⊕ y) ⊕ z = x ⊕ (y ⊕ z);<br />

3. neutral element:<br />

n = C [1, 1, . . ., 1] =<br />

[ 1<br />

D , 1 D , . . ., 1 D]<br />

;<br />

n is the baricenter of the simplex and is unique;

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!