Integral of Positive Simple Function is Well-Defined
Theorem
Let $\struct {X, \Sigma, \mu}$ be a measure space.
Let $f: X \to \R, f \in \EE^+$ be a positive simple function.
Then the $\mu$-integral of $f$, $\map {I_\mu} f$, is well-defined.
That is, for any two standard representations for $f$, say:
- $\ds f = \sum_{i \mathop = 0}^n a_i \chi_{E_i} = \sum_{j \mathop = 0}^m b_j \chi_{F_j}$
it holds that:
- $\ds \sum_{i \mathop = 0}^n a_i \map \mu {E_i} = \sum_{j \mathop = 0}^m b_j \map \mu {F_j}$
Proof
The sets $F_0, \ldots, F_m$ are pairwise disjoint, and:
- $X = \ds \bigcup_{j \mathop = 0}^m F_j$
From Characteristic Function of Disjoint Union, we have:
- $\chi_X = \ds \sum_{j \mathop = 0}^m \chi_{F_j}$
Remark that $\map {\chi_X} x = 1$ for all $x \in X$, so that we have:
\(\ds f\) | \(=\) | \(\ds \sum_{i \mathop = 0}^n a_i \chi_{E_i} \cdot 1\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{i \mathop = 0}^n a_i \chi_{E_i} \paren {\sum_{j \mathop = 0}^m \chi_{F_j} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{i \mathop = 0}^n \sum_{j \mathop = 0}^m a_i \chi_{E_i} \chi_{F_j}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{i \mathop = 0}^n \sum_{j \mathop = 0}^m a_i \chi_{E_i \cap F_j}\) | Characteristic Function of Intersection: Variant 1 |
Repeating the argument with the $E_i$ and $F_j$ interchanged also yields:
- $f = \ds \sum_{j \mathop = 0}^m \sum_{i \mathop = 0}^n b_j \chi_{F_j \cap E_i}$
Now whenever $x \in E_i \cap F_j$, for some $i, j$, then since the $E_i$, $F_j$ are disjoint, we find:
- $x \in E_{i'} \cap F_{j'}$ implies $i = i'$ and $j = j'$
Thus, evaluating both expressions for $\map f x$ we find:
- $a_i = \map f x = b_j$
In conclusion, we have:
- $a_i = b_j$
if $E_i \cap F_j \ne \O$.
Furthermore, we have for all $i$ that:
- $\ds E_i = E_i \cap X = E_i \cap \paren {\bigcup_{j \mathop = 0}^m F_j} = \bigcup_{j \mathop = 0}^m \paren {E_i \cap F_j}$
by Intersection Distributes over Union: General Result.
Similarly, we obtain for all $j$:
- $\ds F_j = F_j \cap X = F_j \cap \paren {\bigcup_{i \mathop = 0}^n E_i} = \bigcup_{i \mathop = 0}^n \paren {F_j \cap E_i}$
With this knowledge, we compute:
\(\ds \sum_{i \mathop = 0}^n a_i \map \mu {E_i}\) | \(=\) | \(\ds \sum_{i \mathop = 0}^n a_i \map \mu {\bigcup_{j \mathop = 0}^m \paren {E_i \cap F_j} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{i \mathop = 0}^n a_i \sum_{j \mathop = 0}^m \map \mu {E_i \cap F_j}\) | Measure is Finitely Additive Function | |||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{i \mathop = 0}^n \sum_{j \mathop = 0}^m a_i \map \mu {E_i \cap F_j}\) | Summation is Linear | |||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{j \mathop = 0}^m \sum_{i \mathop = 0}^n b_j \map \mu {E_i \cap F_j}\) | If $a_i \ne b_j$ then $E_i \cap F_j = \O$; $a_i \cdot 0 = b_j \cdot 0$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{j \mathop = 0}^m b_j \sum_{i \mathop = 0}^n \map \mu {E_i \cap F_j}\) | Summation is Linear | |||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{j \mathop = 0}^m b_j \map \mu {\bigcup_{i \mathop = 0}^n \paren {E_i \cap F_j} }\) | Measure is Finitely Additive Function | |||||||||||
\(\ds \) | \(=\) | \(\ds \sum_{j \mathop = 0}^m b_j \map \mu {F_j}\) |
Hence the result.
$\blacksquare$
Sources
- 2005: René L. Schilling: Measures, Integrals and Martingales ... (previous) ... (next): $9.1$