4 Chapter 1. Introducti<strong>on</strong>
Chapter 2 Compositi<strong>on</strong>al data and their sample space 2.1 Basic c<strong>on</strong>cepts Definiti<strong>on</strong> 2.1. A row vector, x = [x 1 , x 2 , . . .,x D ], is defined as a D-part compositi<strong>on</strong> when all its comp<strong>on</strong>ents are strictly positive real numbers and they carry <strong>on</strong>ly relative informati<strong>on</strong>. Indeed, that compositi<strong>on</strong>al informati<strong>on</strong> is relative is implicitly stated in the units, as they are always parts of a whole, like weight or volume percent, ppm, ppb, or molar proporti<strong>on</strong>s. The most comm<strong>on</strong> examples have a c<strong>on</strong>stant sum κ and are known in the geological literature as closed data (Chayes, 1971). Frequently, κ = 1, which means that measurements have been made in, or transformed to, parts per unit, or κ = 100, for measurements in percent. Other units are possible, like ppm or ppb, which are typical examples for compositi<strong>on</strong>al data where <strong>on</strong>ly a part of the compositi<strong>on</strong> has been recorded; or, as recent studies have shown, even c<strong>on</strong>centrati<strong>on</strong> units (mg/L, meq/L, molarities and molalities), where no c<strong>on</strong>stant sum can be feasibly defined (Buccianti and Pawlowsky-Glahn, 2005; Otero et al., 2005). Definiti<strong>on</strong> 2.2. The sample space of compositi<strong>on</strong>al data is the simplex, defined as S D = {x = [x 1 , x 2 , . . ., x D ] |x i > 0, i = 1, 2, . . ., D; D∑ x i = κ}. (2.1) However, this definiti<strong>on</strong> does not include compositi<strong>on</strong>s in e.g. meq/L. Therefore, a more general definiti<strong>on</strong>, together with its interpretati<strong>on</strong>, is given in Secti<strong>on</strong> 2.2. Definiti<strong>on</strong> 2.3. For any vector of D real positive comp<strong>on</strong>ents z = [z 1 , z 2 , . . .,z D ] ∈ R D + 5 i=1