1 / 116

Maximum Likelihood

Maximum Likelihood. Benjamin Neale Boulder Workshop 2012. We will cover. Easy introduction to probability Rules of probability How to calculate likelihood for discrete outcomes Confidence intervals in likelihood Likelihood for continuous data. Starting simple.

sasson
Download Presentation

Maximum Likelihood

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Maximum Likelihood Benjamin Neale Boulder Workshop 2012

  2. We will cover • Easy introduction to probability • Rules of probability • How to calculate likelihood for discrete outcomes • Confidence intervals in likelihood • Likelihood for continuous data

  3. Starting simple • Let’s think about probability

  4. Starting simple • Let’s think about probability • Coin tosses • Winning the lottery • Roll of the die • Roulette wheel

  5. Starting simple • Let’s think about probability • Coin tosses • Winning the lottery • Roll of the die • Roulette wheel • Chance of an event occurring

  6. Starting simple • Let’s think about probability • Coin tosses • Winning the lottery • Roll of the die • Roulette wheel • Chance of an event occurring • Written as P(event) = probability of the event

  7. Simple probability calculations • To get comfortable with probability, let’s solve these problems: • Probability of rolling an even number on a six-sided die • Probability of pulling a club from a deck of cards

  8. Simple probability calculations • To get comfortable with probability, let’s solve these problems: • Probability of rolling an even number on a six-sided die ½ or 0.5 • Probability of pulling a club from a deck of cards ¼ or 0.25

  9. Simple probability rules • P(A and B) = P(A)*P(B)

  10. Simple probability rules • P(A and B) = P(A)*P(B) • E.g. what is the probability of tossing 2 heads in a row?

  11. Simple probability rules • P(A and B) = P(A)*P(B) • E.g. what is the probability of tossing 2 heads in a row? • A = Heads and B = Heads so,

  12. Simple probability rules • P(A and B) = P(A)*P(B) • E.g. what is the probability of tossing 2 heads in a row? • A = Heads and B = Heads so, • P(A) = ½, P(B) = ½,P(A and B) = ¼

  13. Simple probability rules • P(A and B) = P(A)*P(B) • E.g. what is the probability of tossing 2 heads in a row? • A = Heads and B = Heads so, • P(A) = ½, P(B) = ½, P(A and B) = ¼ *We assume independence

  14. Simple probability rules cnt’d • P(A or B) = P(A) + P(B) – P(A and B)

  15. Simple probability rules cnt’d • P(A or B) = P(A) + P(B) – P(A and B) • What is the probability of rolling a 1 or a 4?

  16. Simple probability rules cnt’d • P(A or B) = P(A) + P(B) – P(A and B) • What is the probability of rolling a 1 or a 4? • A = rolling a 1 and B = rolling a 4

  17. Simple probability rules cnt’d • P(A or B) = P(A) + P(B) – P(A and B) • What is the probability of rolling a 1 or a 4? • A = rolling a 1 and B = rolling a 4 • P(A) = , P(B) = , P(A or B) = 1 1 1 6 6 3

  18. Simple probability rules cnt’d • P(A or B) = P(A) + P(B) – P(A and B) • What is the probability of rolling a 1 or a 4? • A = rolling a 1 and B = rolling a 4 • P(A) = , P(B) = , P(A or B) = 1 1 1 6 6 3 *We assume independence

  19. Recap of rules • P(A and B) = P(A)*P(B) • P(A or B) = P(A) + P(B) – P(A and B) • Sometimes things are ‘exclusive’ such as rolling a 6 and rolling a 4. It cannot occur in the same trial implies P(A and B) = 0 Assuming independence

  20. Conditional probabilities • P(X | Y) = the probability of X occurring given Y.

  21. Conditional probabilities • P(X | Y) = the probability of X occurring given Y. • Y can be another event (perhaps that predicts X)

  22. Conditional probabilities • P(X | Y) = the probability of X occurring given Y. • Y can be another event (perhaps that predicts X) • Y can be a probability or set of probabilities

  23. Conditional probabilities • Roll two dice in succession

  24. 1 12 Conditional probabilities • Roll two dice in succession • P(total = 10) =

  25. 1 12 Conditional probabilities • Roll two dice in succession • P(total = 10) = • What is P(total = 10 | 1st die = 5)?

  26. 1 1 12 6 Conditional probabilities • Roll two dice in succession • P(total = 10) = • What is P(total = 10 | 1st die = 5)? • P(total = 10 | 1st die = 5) =

  27. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome:

  28. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome:

  29. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome: Probability # of k results # of trials

  30. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome: Number of combinations of n choose k ! = factorial; n! = n*(n-1)*(n-2)*…*2*1 and factorials are bad for big numbers

  31. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome: Probability of not k occurring Probability of k occurring

  32. Binomial probabilities • Used for two conditions such as coin toss • Determine the chance of any outcome: Probability of not k occurring Number of combinations of n choose k Probability # of positive results Probability of k occurring # of trials ! = factorial; n! = n*(n-1)*(n-2)*…*2*1 and factorials are bad for big numbers

  33. Combinations piece long way • Does it work? Let’s try: How many combinations for 3 heads out of 5 tosses?

  34. Combinations piece long way • Does it work? Let’s try: How many combinations for 3 heads out of 5 tosses? • HHHTT, HHTHT, HHTTH, HTHHT, HTHTH, HTTHH, THHHT, THHTH, THTHH, TTHHH = 10 possible combinations

  35. Combinations piece formula • Does it work? Let’s try: How many combinations for 3 heads out of 5 tosses? • We have 5 choose 3 = 5!/(3!)*(2!) • =(5*4)/2 • =10

  36. Probability roundup • We assumed the ‘true’ parameter values • E.g. P(Heads) = P(Tails) = ½

  37. Probability roundup • We assumed the ‘true’ parameter values • E.g. P(Heads) = P(Tails) = ½ • What happens if we have data and want to determine the parameter values?

  38. Probability roundup • We assumed the ‘true’ parameter values • E.g. P(Heads) = P(Tails) = ½ • What happens if we have data and want to determine the parameter values? • Likelihood works the other way round: what is the probability of the observed data given parameter values?

  39. Concrete example • Likelihood aims to calculate the range of probabilities for observed data, assuming different parameter values.

  40. Concrete example • Likelihood aims to calculate the range of probabilities for observed data, assuming different parameter values. • The set of probabilities is referred to as a likelihood surface

  41. Concrete example • Likelihood aims to calculate the range of probabilities for observed data, assuming different parameter values. • The set of probabilities is referred to as a likelihood surface • We’re going to generate the likelihood surface for a coin tossing experiment

  42. Concrete example • Likelihood aims to calculate the range of probabilities for observed data, assuming different parameter values. • The set of probabilities is referred to as a likelihood surface • We’re going to generate the likelihood surface for a coin tossing experiment • The set of parameter values with the best probability is the maximum likelihood

  43. Coin tossing • I tossed a coin 10 times and get 4 heads and 6 tails

  44. Coin tossing • I tossed a coin 10 times and get 4 heads and 6 tails • From this data, what does likelihood estimate the chance of heads and tails for this coin?

  45. Coin tossing • I tossed a coin 10 times and get 4 heads and 6 tails • From this data, what does likelihood estimate the chance of heads and tails for this coin? • We’re going to calculate: • P(4 heads out of 10 tosses| P(H) = *) • where star takes on a range of values

  46. Calculations • P(4 heads out of 10 tosses | P(H)=0.1) = We can make this easier, as it will be constant C* across all calculations

  47. Calculations • P(4 heads out of 10 tosses | P(H)=0.1) = Now all we do is change the values of p and q

  48. Calculations • P(4 heads out of 10 tosses | P(H)=0.1) =

  49. Table of likelihoods

  50. Table of likelihoods Largest probability observed = maximum likelihood

More Related