• conditional probability : the probability of event once we know happened; becomes the new sample space
    • rearranged it is the general multiplication rule ( Probability Rules ); flipped it is Bayes Rule
    • independent exactly when , and then the intersection is the product
      • two aces with replacement , without replacement
      • independent: roulette spun twice, a coin flipped twice, two dice, sampling with replacement
  • example: phones make zero ( ), one ( ) or more than one ( ) handoffs as they move from cell to cell, and calls are long ( , over 3 minutes ) or brief ( ); 100 calls. Are length and handoffs independent?
H0H1H2Total
L10102040
B40101060
Total502030n = 100
  • marginals: , ; , ,
  • conditionals, divide each row by its total: and
givenH0H1H2
L0.250.250.50
B0.670.170.17
  • or each column: , ,
    • conditionals similar to the marginals means no association, independent; different means associated, not independent Screenshot 2024-09-12 at 10.44.38 AM.png
    • one cell is enough to show it: but , so not independent
  • tree diagrams: 10 balls in an urn, 3 orange, 5 red, 2 blue, draw two without replacement; each branch is conditional on the one before it
graph LR
    Start["Start (10 balls)"]

    Start --> O1["Orange (3/10)"]
    Start --> R1["Red (5/10)"]
    Start --> B1["Blue (2/10)"]

    O1 --> O2["Orange (2/9)"]
    O1 --> R2["Red (5/9)"]
    O1 --> B2["Blue (2/9)"]

    R1 --> O3["Orange (3/9)"]
    R1 --> R3["Red (4/9)"]
    R1 --> B3["Blue (2/9)"]

    B1 --> O4["Orange (3/9)"]
    B1 --> R4["Red (5/9)"]
    B1 --> B4["Blue (1/9)"]
  • multiply along a path for a joint probability:
    • add the paths that end in orange for the second draw:
    • same as the first draw: the draws are dependent, but each draw on its own still has the original odds