-
Notifications
You must be signed in to change notification settings - Fork 62
/
02-probability.Rmd
296 lines (184 loc) · 11.4 KB
/
02-probability.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
# Probability Review
We give a very brief review of some necessary probability concepts. As the treatment is less than complete, a list of references is given at the end of the chapter. For example, we ignore the usual recap of basic set theory and omit proofs and examples.
## Probability Models
When discussing probability models, we speak of random **experiments** that produce one of a number of possible **outcomes**.
A **probability model** that describes the uncertainty of an experiment consists of two elements:
- The **sample space**, often denoted as $\Omega$, which is a set that contains all possible outcomes.
- A **probability function** that assigns to an event $A$ a non-negative number, $P[A]$, that represents how likely it is that event $A$ occurs as a result of the experiment.
We call $P[A]$ the **probability** of event $A$. An **event** $A$ could be any subset of the sample space, not necessarily a single possible outcome. The probability law must follow a number of rules, which are the result of a set of axioms that we introduce now.
## Probability Axioms
Given a sample space $\Omega$ for a particular experiment, the **probability function** associated with the experiment must satisfy the following axioms.
1. *Non-negativity*: $P[A] \geq 0$ for any event $A \subset \Omega$.
2. *Normalization*: $P[\Omega] = 1$. That is, the probability of the entire space is 1.
3. *Additivity*: For mutually exclusive events $E_1, E_2, \ldots$
$$
P\left[\bigcup_{i = 1}^{\infty} E_i\right] = \sum_{i = 1}^{\infty} P[E_i]
$$
Using these axioms, many additional probability rules can easily be derived.
## Probability Rules
Given an event $A$, and its complement, $A^c$, that is, the outcomes in $\Omega$ which are not in $A$, we have the **complement rule**:
$$
P[A^c] = 1 - P[A]
$$
In general, for two events $A$ and $B$, we have the **addition rule**:
$$
P[A \cup B] = P[A] + P[B] - P[A \cap B]
$$
If $A$ and $B$ are also *disjoint*, then we have:
$$
P[A \cup B] = P[A] + P[B]
$$
If we have $n$ mutually exclusive events, $E_1, E_2, \ldots E_n$, then we have:
$$
P\left[\textstyle\bigcup_{i = 1}^{n} E_i\right] = \sum_{i = 1}^{n} P[E_i]
$$
Often, we would like to understand the probability of an event $A$, given some information about the outcome of event $B$. In that case, we have the **conditional probability rule** provided $P[B] > 0$.
$$
P[A \mid B] = \frac{P[A \cap B]}{P[B]}
$$
Rearranging the conditional probability rule, we obtain the **multiplication rule**:
$$
P[A \cap B] = P[B] \cdot P[A \mid B] \cdot
$$
For a number of events $E_1, E_2, \ldots E_n$, the multiplication rule can be expanded into the **chain rule**:
$$
P\left[\textstyle\bigcap_{i = 1}^{n} E_i\right] = P[E_1] \cdot P[E_2 \mid E_1] \cdot P[E_3 \mid E_1 \cap E_2] \cdots P\left[E_n \mid \textstyle\bigcap_{i = 1}^{n - 1} E_i\right]
$$
Define a **partition** of a sample space $\Omega$ to be a set of disjoint events $A_1, A_2, \ldots, A_n$ whose union is the sample space $\Omega$. That is
$$
A_i \cap A_j = \emptyset
$$
for all $i \neq j$, and
$$
\bigcup_{i = 1}^{n} A_i = \Omega.
$$
Now, let $A_1, A_2, \ldots, A_n$ form a partition of the sample space where $P[A_i] > 0$ for all $i$. Then for any event $B$ with $P[B] > 0$ we have **Bayes' Rule**:
$$
P[A_i | B] = \frac{P[A_i]P[B | A_i]}{P[B]} = \frac{P[A_i]P[B | A_i]}{\sum_{i = 1}^{n}P[A_i]P[B | A_i]}
$$
The denominator of the latter equality is often called the **law of total probability**:
$$
P[B] = \sum_{i = 1}^{n}P[A_i]P[B | A_i]
$$
Two events $A$ and $B$ are said to be **independent** if they satisfy
$$
P[A \cap B] = P[A] \cdot P[B]
$$
This becomes the new multiplication rule for independent events.
A collection of events $E_1, E_2, \ldots E_n$ is said to be independent if
$$
P\left[\bigcap_{i \in S} E_i \right] = \prod_{i \in S}P[E_i]
$$
for every subset $S$ of $\{1, 2, \ldots n\}$.
If this is the case, then the chain rule is greatly simplified to:
$$
P\left[\textstyle\bigcap_{i = 1}^{n} E_i\right] = \prod_{i=1}^{n}P[E_i]
$$
## Random Variables
A **random variable** is simply a *function* which maps outcomes in the sample space to real numbers.
### Distributions
We often talk about the **distribution** of a random variable, which can be thought of as:
$$
\text{distribution} = \text{list of possible} \textbf{ values} + \text{associated} \textbf{ probabilities}
$$
This is not a strict mathematical definition, but is useful for conveying the idea.
If the possible values of a random variables are *discrete*, it is called a *discrete random variable*. If the possible values of a random variables are *continuous*, it is called a *continuous random variable*.
### Discrete Random Variables
The distribution of a discrete random variable $X$ is most often specified by a list of possible values and a probability **mass** function, $p(x)$. The mass function directly gives probabilities, that is,
$$
p(x) = p_X(x) = P[X = x].
$$
Note we almost always drop the subscript from the more correct $p_X(x)$ and simply refer to $p(x)$. The relevant random variable is discerned from context
The most common example of a discrete random variable is a **binomial** random variable. The mass function of a binomial random variable $X$, is given by
$$
p(x | n, p) = {n \choose x} p^x(1 - p)^{n - x}, \ \ \ x = 0, 1, \ldots, n, \ n \in \mathbb{N}, \ 0 < p < 1.
$$
This line conveys a large amount of information.
- The function $p(x | n, p)$ is the mass function. It is a function of $x$, the possible values of the random variable $X$. It is conditional on the **parameters** $n$ and $p$. Different values of these parameters specify different binomial distributions.
- $x = 0, 1, \ldots, n$ indicates the **sample space**, that is, the possible values of the random variable.
- $n \in \mathbb{N}$ and $0 < p < 1$ specify the **parameter spaces**. These are the possible values of the parameters that give a valid binomial distribution.
Often all of this information is simply encoded by writing
$$
X \sim \text{bin}(n, p).
$$
### Continuous Random Variables
The distribution of a continuous random variable $X$ is most often specified by a set of possible values and a probability **density** function, $f(x)$. (A cumulative density or moment generating function would also suffice.)
The probability of the event $a < X < b$ is calculated as
$$
P[a < X < b] = \int_{a}^{b} f(x)dx.
$$
Note that densities are **not** probabilities.
The most common example of a continuous random variable is a **normal** random variable. The density of a normal random variable $X$, is given by
$$
f(x | \mu, \sigma^2) = \frac{1}{\sigma\sqrt{2\pi}} \cdot \exp\left[\frac{-1}{2} \left(\frac{x - \mu}{\sigma}\right)^2 \right], \ \ \ -\infty < x < \infty, \ -\infty < \mu < \infty, \ \sigma > 0.
$$
- The function $f(x | \mu, \sigma^2)$ is the density function. It is a function of $x$, the possible values of the random variable $X$. It is conditional on the **parameters** $\mu$ and $\sigma^2$. Different values of these parameters specify different normal distributions.
- $-\infty < x < \infty$ indicates the sample space. In this case, the random variable may take any value on the real line.
- $-\infty < \mu < \infty$ and $\sigma > 0$ specify the parameter space. These are the possible values of the parameters that give a valid normal distribution.
Often all of this information is simply encoded by writing
$$
X \sim N(\mu, \sigma^2)
$$
### Several Random Variables
Consider two random variables $X$ and $Y$. We say they are independent if
$$
f(x, y) = f(x) \cdot f(y)
$$
for all $x$ and $y$. Here $f(x, y)$ is the **joint** density (mass) function of $X$ and $Y$. We call $f(x)$ the **marginal** density (mass) function of $X$. Then $f(y)$ the marginal density (mass) function of $Y$. The joint density (mass) function $f(x, y)$ together with the possible $(x, y)$ values specify the joint distribution of $X$ and $Y$.
Similar notions exist for more than two variables.
## Expectations
For discrete random variables, we define the **expectation** of the function of a random variable $X$ as follows.
$$
\mathbb{E}[g(X)] \triangleq \sum_{x} g(x)p(x)
$$
For continuous random variables we have a similar definition.
$$
\mathbb{E}[g(X)] \triangleq \int g(x)f(x) dx
$$
For specific functions $g$, expectations are given names.
The **mean** of a random variable $X$ is given by
$$
\mu_{X} = \text{mean}[X] \triangleq \mathbb{E}[X].
$$
So for a discrete random variable, we would have
$$
\text{mean}[X] = \sum_{x} x \cdot p(x)
$$
For a continuous random variable we would simply replace the sum by an integral.
The **variance** of a random variable $X$ is given by
$$
\sigma^2_{X} = \text{var}[X] \triangleq \mathbb{E}[(X - \mathbb{E}[X])^2] = \mathbb{E}[X^2] - (\mathbb{E}[X])^2.
$$
The **standard deviation** of a random variable $X$ is given by
$$
\sigma_{X} = \text{sd}[X] \triangleq \sqrt{\sigma^2_{X}} = \sqrt{\text{var}[X]}.
$$
The **covariance** of random variables $X$ and $Y$ is given by
$$
\text{cov}[X, Y] \triangleq \mathbb{E}[(X - \mathbb{E}[X])(Y - \mathbb{E}[Y])] = \mathbb{E}[XY] - \mathbb{E}[X] \cdot \mathbb{E}[Y].
$$
## Likelihood
Consider $n$ iid random variables $X_1, X_2, \ldots X_n$. We can then write their **likelihood** as
$$
\mathcal{L}(\theta \mid x_1, x_2, \ldots x_n) = \prod_{i = i}^n f(x_i; \theta)
$$
where $f(x_i; \theta)$ is the density (or mass) function of random variable $X_i$ evaluated at $x_i$ with parameter $\theta$.
Whereas a probability is a function of a possible observed value given a particular parameter value, a likelihood is the opposite. It is a function of a possible parameter value given observed data.
Maximumizing likelihood is a common techinque for fitting a model to data.
## Videos
The YouTube channel [mathematicalmonk](https://www.youtube.com/user/mathematicalmonk) has a great [Probability Primer playlist](https://www.youtube.com/playlist?list=PL17567A1A3F5DB5E4) containing lectures on many fundamental probability concepts. Some of the more important concepts are covered in the following videos:
- [Conditional Probability](https://www.youtube.com/watch?v=5BWk5qe5EJ8&index=11&list=PL17567A1A3F5DB5E4)
- [Independence](https://www.youtube.com/watch?v=KK9jvGl9FY0&index=12&list=PL17567A1A3F5DB5E4)
- [More Independence](https://www.youtube.com/watch?v=RMS-WglZP-c&index=13&list=PL17567A1A3F5DB5E4)
- [Bayes Rule](https://www.youtube.com/watch?v=cM1BqBv11U8&index=14&list=PL17567A1A3F5DB5E4)
## References
Any of the following are either dedicated to, or contain a good coverage of the details of the topics above.
- Probability Texts
- [Introduction to Probability](http://athenasc.com/probbook.html) by Dimitri P. Bertsekas and John N. Tsitsiklis
- [A First Course in Probability](https://www.pearsonhighered.com/program/Ross-First-Course-in-Probability-A-9th-Edition/PGM110742.html) by Sheldon Ross
- Machine Learning Texts with Probability Focus
- [Probability for Statistics and Machine Learning](http://www.springer.com/us/book/9781441996336) by Anirban DasGupta
- [Machine Learning: A Probabilistic Perspective](https://mitpress.mit.edu/books/machine-learning-0) by Kevin P. Murphy
- Statistics Texts with Introduction to Probability
- [Probability and Statistical Inference](https://www.pearsonhighered.com/program/Hogg-Probability-and-Statistical-Inference-9th-Edition/PGM91556.html) by Robert V. Hogg, Elliot Tanis, and Dale Zimmerman
- [Introduction to Mathematical Statistics](https://www.pearsonhighered.com/program/Hogg-Introduction-to-Mathematical-Statistics-7th-Edition/PGM49624.html) by Robert V. Hogg, Joseph McKean, and Allen T. Craig