Probability and Stochastics (Graduate Texts in Mathematics 261)

Erhan Cinlar

Chapter 1: Measure and Integration

Exercises

1.9   Partition generated σ\sigma-algebras.
a)\quad \text{a)} \> Let C={A,B,C}\mathcal{C} = \{A, B, C\} be a partition of EE. List the elements of σC\sigma \mathcal{C}.
Solution
We start off with the elements that are members of all σ\sigma-algebras
,E\emptyset, E
Then we go on to list the elements of C\mathcal{C}
A,B,CA, B, C
We also need the complements of each of those
BC,AC,ABB \cup C, A \cup C, A \cup B
where we are using the fact that C\mathcal{C} is a partition to deduce that EA=BCE \setminus A = B \cup C, etc. I would go on to list their unions but I would just be repeating elements (because ABC=EA \cup B \cup C = E). So far what I have is
σC={,A,B,C,BC,AC,AB,E}\sigma \mathcal{C} = \{\emptyset, A, B, C, B \cup C, A \cup C, A \cup B, E\}
This should be correct as it is closed under finite unions and complements, and is the smallest such algebra containing C\mathcal{C}. \square


b)\quad \text{b)} \> Let C\mathcal{C} be a (countable) partition of EE. Show that every element of σC\sigma \mathcal{C} is a countable union of elements taken from C\mathcal{C}. Hint: Let E\mathcal{E} be the collection of all sets that are countable unions of elements taken from C\mathcal{C}. Show that E\mathcal{E} is a σ\sigma-algebra, and argue that E=σC\mathcal{E} = \sigma \mathcal{C}.
Solution
First let's use the fact that C\mathcal{C} is a countable partition of EE, so we have C={Ci:iI}\mathcal{C} = \{C_i: i \in I\}, such that iICi=E\bigcup_{i \in I} C_i = E and ij    CiCj=i \neq j \implies C_i \cap C_j = \emptyset. Let E\mathcal{E} be the collection of all sets that are countable unions of elements taken from C\mathcal{C}. We proceed by showing that E\mathcal{E} is a σ\sigma-algebra.
Let AEA \in \mathcal{E}, so for some JIJ \subseteq I we have A=jJCjA = \bigcup_{j \in J}C_j. Let B=kI,kJCkB = \bigcup_{k \in I, k \notin J} C_k. Clearly BEB \in \mathcal{E}, and propose that EA=BE \setminus A = B. Let xEAx \in E \setminus A, since xEx \in E and iICi=E\bigcup_{i \in I} C_i = E we have some iIi \in I such that xCix \in C_i, and since xAx \notin A we have iJi \notin J, so by our definition of BB we have xBx \in B. Alternatively, if xBx \in B and xEAx \notin E \setminus A then xAB=x \in A \cap B = \emptyset, so we can not have this. We justify AB=A \cap B = \emptyset as follows: let xABx \in A \cap B. Then for some jJj \in J we have xCjx \in C_j and for some kJk \notin J we have xCkx \in C_k. Since C\mathcal{C} is a partition and kjk \neq j we have CjCk=C_j \cap C_k = \emptyset. Anyways, since EA=BEE \setminus A = B \in \mathcal{E} for all AEA \in \mathcal{E}, we have shown 1.3 a)\text{1.3 a)}.
Showing 1.3 b)\text{1.3 b)} is trivial, as the countable union of a countable union is a countable union.
Combining 1.3 a)\text{1.3 a)} and 1.3 b)\text{1.3 b)} shows that E\mathcal{E} is a σ\sigma-algebra.
To finish the proof we need to show that E=σC\mathcal{E} = \sigma \mathcal{C}. It is obvious that E\mathcal{E} contains C\mathcal{C} and we know that E\mathcal{E} is a σ\sigma-algebra so it is obvious that σCE\sigma \mathcal{C} \subseteq \mathcal{E}. Because a σ\sigma-algebra is by definition closed under countable unions we must also have EσC\mathcal{E} \subseteq \sigma \mathcal{C}, so E=σC\mathcal{E} = \sigma \mathcal{C}. \square


c)\quad \text{c)} \> Let E=RE = \R, the set of all real numbers. Let C\mathcal{C} be the collection of all singleton subsets of R\R, that is, each element of C\mathcal{C} is a set that consists of exactly one point in R\R. Show that every element of σC\sigma \mathcal{C} is either a countable set or the complement of a countable set. Incidentally, σC\sigma \mathcal{C} is much smaller than B(R)\mathcal{B}(\R); for instance, the interval (0,1)(0, 1) belongs to the latter but not to the former.
Solution
Let E\mathcal{E} be the set of all finite unions of singletons and their complements. We propose that σC=E\sigma \mathcal{C} = \mathcal{E}.
First we show that E\mathcal{E} is indeed a σ\sigma-algebra. By definition this means showing that

Anyways after having showed that E\mathcal{E} is a σ\sigma-algebra the rest of this problem is trivial. CE\mathcal{C} \subset \mathcal{E} is obvious by construction, so by the definition of a σ\sigma-algebra being generated we have σCE\sigma \mathcal{C} \subset \mathcal{E}. EσC\mathcal{E} \subset \sigma \mathcal{C} is trivial because σ\sigma-algebras are closed under finite unions and complements. \quad \square


1.10   Comparisons. Let C\mathcal{C} and D\mathcal{D} be two collections of subsets of EE. Show the following:
a)\quad \text{a)} \quad If CD\mathcal{C} \subset \mathcal{D} then σCσD\sigma \mathcal{C} \subset \sigma \mathcal{D}
b)\quad \text{b)} \quad If CσD\mathcal{C} \subset \sigma \mathcal{D} then σCσD\sigma C \subset \sigma D
c)\quad \text{c)} \quad If CσDC \subset \sigma \mathcal{D} and DσC\mathcal{D} \subset \sigma \mathcal{C}, then σC=σD\sigma \mathcal{C} = \sigma \mathcal{D}
d)\quad \text{d)} \quad If CDσC\mathcal{C} \subset \mathcal{D} \subset \sigma \mathcal{C}, then σC=σD\sigma \mathcal{C} = \sigma \mathcal{D}


1.11   Borel σ\sigma-algebra on R\R. Every open subset of R=(,+)\R = (-\infty, +\infty), the real line, is a countable union of open intervals. Use this fact to show that BR\mathcal{B}_\R is generated by the collection of all open intervals.


1.12   Continuation. Show that every interval of R\R is a Borel set. In particular, (,x)(-\infty, x), (,x](-\infty,x], (x,y](x,y], [x,y][x,y] are all Borel sets. For each xx, the singleton {x}\{x\} is a Borel set.


1.13   Continuation. Show that BR\mathcal{B}_\R is also generated by any one of the following (and many others):
a)\quad \text{a)} \quad The collection of all intervals of the form (,x](-\infty, x].
b)\quad \text{b)} \quad The collection of all intervals of the form (x,y](x,y].
c)\quad \text{c)} \quad The collection of all intervals of the form [x,y][x,y].
d)\quad \text{d)} \quad The collection of all intervals of the form (x,)(x, \infty).

Moreover, in each case, xx and yy can be limited to rationals.


1.14   Lemma 1.7. Prove.

1.7   Lemma. Let D\mathcal{D} be a d-system on EE. Fix DD in D\mathcal{D} and let

D^={AD:ADD}\hat{\mathcal{D}} = \{A \in \mathcal{D}: A \cap D \in \mathcal{D}\}

Then, D^\hat{\mathcal{D}} is again a d-system.


1.15   Trace spaces. Let (E,E)(E, \mathcal{E}) be a measurable space. Fix DED \subset E and let

D=ED={AD:AE}.\mathcal{D} = \mathcal{E} \cap D = \{A \cap D: A \in \mathcal{E}\}.

Show that D\mathcal{D} is a σ\sigma-algebra on DD. It is called the trace on E\mathcal{E} on DD, and (D,D)(D,\mathcal{D}) is called the trace of (E,E)(E,\mathcal{E}) on DD.

\\

Probability and Stochastics (Graduate Texts in Mathematics 261)

Chapter 1: Measure and Integration