Lecture Notes on Compositional Data Analysis - Sedimentology ...
Lecture Notes on Compositional Data Analysis - Sedimentology ...
Lecture Notes on Compositional Data Analysis - Sedimentology ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
14 Chapter 3. Aitchis<strong>on</strong> geometry<br />
3.3 Inner product, norm and distance<br />
To obtain a linear vector space structure, we take the following inner product, with<br />
associated norm and distance:<br />
Definiti<strong>on</strong> 3.3. Inner product of x,y ∈ S D ,<br />
〈x,y〉 a = 1 D∑ D∑<br />
ln x i<br />
ln y i<br />
.<br />
2D x j y j<br />
Definiti<strong>on</strong> 3.4. Norm of x ∈ S D ,<br />
‖x‖ a<br />
= √ 1<br />
2D<br />
i=1<br />
D∑<br />
i=1<br />
j=1<br />
D∑<br />
j=1<br />
i=1<br />
(<br />
ln x i<br />
x j<br />
) 2<br />
.<br />
Definiti<strong>on</strong> 3.5. Distance between x and y ∈ S D ,<br />
d a (x,y) = ‖x ⊖ x‖ a<br />
= √ 1 D∑ D∑<br />
(<br />
ln x i<br />
− ln y ) 2<br />
i<br />
.<br />
2D x j y j<br />
In practice, alternative but equivalent expressi<strong>on</strong>s of the inner product, norm and<br />
distance may be useful. Two possible alternatives of the inner product follow:<br />
(<br />
〈x,y〉 a = 1 D−1<br />
∑ D∑<br />
ln x i<br />
ln y D∑<br />
i<br />
= ln x i ln x j − 1 D<br />
) (<br />
∑ D<br />
)<br />
∑<br />
ln x j ln x k ,<br />
D x j y j D<br />
i=1 j=i+1<br />
i