Week 47

Measures and Integrals

Monday Lecture

In the first lecture we introduce how probablity theory and the modeling of experiments and observations can be phrased using abstract measure theory (Chapter 13). A probability measure is simply a measure whose total mass is 1. Probability theory can thus be viewed as a special case of general measure theory, but it gets its distinct flavor from the interpretations. We present the frequentistic and the Bayesian interpretations of probability measures. A link between the probabilistic concept of expectations and the concept of integrals is established.

We also introduce the transformations or image measures, Section 10.1, which will play a central role. As we will see, this construction is the key to understanding how probability models transform either by our computations with data or by our limited ability to make observations.

The ContaminatedSoil case (not in the book) on quantitative risk assessment and decision making based on probability theory is introduced, and this case will follow us throughout the course as we develop the necessary theory to gradually solve questions and expand on various aspects of the case.

To prepare yourself you should read Chapter 2 as a brush up on the fundamental concepts and results that were taught on An2. We will only briefly cover aspects of this chapter in the lecture, but you will also work with Chapter 2 material in the exercise class.

In addition, it will be a really good idea to consult the appendices A, B and C for background material on sets, countability and properties of the real numbers that we will use in the course.
Monday Exercises
2.2, 2.3, 2.7, 2.22
2.22 is relatively technical, but it is a quite good exercise for training fundamental concepts like nullsets and sigma-algebras.
Wednesday Lectures

Expectations are integrals, and we show the abstract result on integration w.r.t. image measures in Section 10.3. The image of a probability measure is the probability distribution obtained by a transformation, e.g. a squaring, of the observable. Theorem 10.8 gives the abstract formula for how integrals and measures transform.

Probability measures, and other measures for that matter, are often given in terms of a density (Chapter 11). Typically in applications the density is w.r.t. the Lebesgue measure. In this case many practical computations with probability measures can be carried out by more or less standard Riemann integration techniques.

In the lectures we develop the theory for integration w.r.t. a measure given by a density (Chapter 11) and how we transform measures given by a density w.r.t. the Lebesgue measure (Section 12.2).

We skip Lemma 11.5 but Lemma 12.4 (in Section 12.1) is also covered this week.

The first assignment is made available on Absalon.
Wednesday Exercises
10.1, 10.2, 10.3, 13.1, 13.2, 13.3