Bayes in Practice & Random Variables stat 430 Heike Hofmann Outline • Bernoulli Experiments • Discrete Random Variables • Expected Value Bayes Tree Diagrams Visualization of Total Probability 24 The probability of a nodeCHAPTER is given as the product of all B1 P(A| B1) A B1 probabilities along the edges P(B ) B2 P(A| B2) back to the root (Definition A B2 P(B ) of conditional probability) 1 2 P(Bk) Bk P(A| Bk) cover A Bk The probability of an event The probability of each is the sum of node the in the tree ca tiplying all probabilities from the root to th probabilities of all final diagrams). Summing up all the involved probabilities in the l nodes (leaves) rule). If the set B1 , . . . , Bk is a cover of the sample space Ω for an event A by (cf. fig.??): Bayes’ Rule • Bayes Theorem P (A) = k � i=1 P (Bi ) · P (A|B Let B1 , B2 , B3 , ... be a cover of the sample space, then is a cover of the sample space Ω, we can compute the probab P (A|Bj ) · P (Bj ) g.??): P (B |A) = P (Bj ∩ A) = �k j P (A) i=1 P (A|Bi ) · P (Bi ) P (A) = k � i=1 P (Bi ) · P (A|Bi ). ∩ A) P (A|Bj ) · P (Bj ) = �k A) for all j and ∅ = � A ⊂ Ω. Example: Forensic Analysis Setup: DNA is found at a crime scene Probability for a (random) DNA match is I in10 Mio DNA test procedure is sometimes faulty: false negative: P(test neg | match) = 0.0000001 false positive: P(test pos | no match) = 0.000001 Assume, police has a positive test result from a suspect. What is the probability that they have found the perpetrator? Example: Monty Hall Problem Setup: - Two goats and one car behind doors - Game show contestant has to pick a door (but doesn’t open it yet) - Host reveals one of the goats - Final choice for contestant: stick with original pick or switch to the other door? What is the probability of winning the car? Bernoulli & Binomial Bernoulli Experiments • Outcome: success or failure; 0 or 1 • P(success) = p, P(failure) = 1 - p • sequence of (independent) repetitions: sequence of Bernoulli experiments X function has two main properties: Properties of a pmf Sequence of Bernoulli Experiments must be between 0 and 1 0 ≤ p (x) ≤ 1 for all x ∈ {x X � all values is 1 i pX (xi ) = 1 k repetitions of Bernoulli experiment: � E[h(X)] = space h(x · pX (xi )of=:k-digit µ Write sample asi )sequence binary numbers:i • • Ωk = { 00...00, 00...01, 00...10, 00...11, ..., 11...00, 11...01, 11...10, 11...11} X : Ω �→ R is called a random variable. 1, x ll values is 1 � i pX (xi ) =1 Probability assignment � E[h(X)] = h(x ) · p (x ) =: µ in Bernoulli Spaces i X i i • If experiments are independent: Ωk = {For sequence 00...00, s00...01, 00...10, 00...11, in sample space • ..., P(s) = pi (1-p)k-i if s has11...00, i successes and 11...10, k-i failures. 11...01, 11...11} (Hint: substitute 0s by 1-p and 1s by p) Ω �→ R is called a random variable. # of sequences with exactly i successes is � � k i � � n • i Ωk = { 00...00, 00...01, 00...10, 00...11, Binomial distribution ..., 11...00, 11...01, 11...10, 11...11} • X : Ω �→ R Let is called a random variable. X be the number of successes in n independent Bernoulli � �experiments with P k (success)=p, i then � � n k P (X = k) = p (1 − p)n−k k (i) • 0 ≤ P (A) ≤ 1 Ωk = { 00...00, 00...01, 0 Random Variables ..., 11...00, 11...01, 1 • Definition: AAfunction calleda arandom random variab function X : Ω �→ R is called variable � � k image of X: i im(X) = X(Omega) = set of all possible � � values X can take n k P (X = k) = p (1 k Examples: • #heads in 10 throws, #of songs from 80s in 1h of LITE 104.1 (or KURE 88.5 FM), (i) 0 ≤ P (A) ≤ 1 winnings in Darts (ii) P (∅) = 0 Discrete R.V.s • If the image of a random variable is finite (or countable infinite), the random variable is a discrete variable • probability mass function Probability mass function � 2 � V ar[X] = (x i− (pmf) VE[X]) ar[X] =· pX (x (xi ) − E[X]) i i i 2 · pX (x) :=function P (X = px) is called the probability mass func The X (x) := P (X = x) is called the pro unction has mass two main properties: of a pmf p probability function has two Properties main properties: Pro X, if and only if pX is a pmf, iff Theorem: ust(i) be all between 1 01between ≤ 1 for ∈(x) {x1≤ , x12 X (x) values be 0≤im(X) and 1 0all≤xpX 0 ≤0must pand forpall x in X(x) ≤ � � ll (ii) values 1 ofi pall = {x1, x2, ... } i ) = 1is 1for im(X) X (x theissum values i pX (xi ) = 1 • • • E[h(X)] = � i � h(xE[h(X)] ) =: µh(xi ) · pX (xi i ) · pX (xi= i Expectation y if Expected Value es must be between 0 and � 1 0 ≤ pX (x) ≤ 1 for all x ∈ { 2 V ar[X] (xi − E[X]) · pX (xi ) �= m of all values is 1 i pXi(xi ) = 1 n pX (x) := (X = x)value is called the probability TheP expected of random variable X ismass fu � that we Properties ss functionthe haslong twoterm main properties: of a pm average will see, E[h(X)] = h(xi ) · pX (xi ) =: µ when we repeat the same experiment over if i and over: must be between 0 and 1 0� ≤ pX (x) ≤ 1 for all x ∈ {x1 , xi · pX (xi ) =: µ �E[X] = of all values is 1 i pX (xi ) =i 1 for additional function h, we get: � =00...00, h(x (xi ) =: 00...11, µ ΩE[h(X)] 00...01, i ) · pX 00...10, k ={ • • ..., i