NAME: Data from the game "Pass the Pigs" TYPE: Observational SIZE: 6000 Observations, 5 Variables DESCRIPTIVE ABSTRACT: This dataset contains information collected from rolling the pair of pigs (found in the game "Pass the Pigs") 6000 times. A description of the rules, scoring configurations, and data collection method are included in the accompanying paper. BRIEF DESCRIPTION of "PASS THE PIGS": A player rolls two pig-shaped rubber dice and earns or loses points depending on the configuration of the rolled pigs. Players compete individually to earn 100 points. VARIABLE DESCRIPTIONS: Each row of the data set represents information collected on a single roll of the pigs. There are 6000 rows, and 6 space-delimited columns. -COLUMN 1 (roll): The roll number. Values are unique integers 1 through 6000. -COLUMN 2 (black): The position number of the marked (black) pig. Values are integers from 1 to 7: 1 = Dot Up 2 = Dot Down 3 = Trotter 4 = Razorback 5 = Snouter 6 = Leaning Jowler 7 = Pigs are touching one another -COLUMN 3 (pink): The position number of the unmarked (pink) pig. Values are the integers from 1 to 7 as above. -COLUMN 4 (score): The score of the roll. All scores are face-value from the ten integer set {-1, 0, 1, 5, 10, 15, 20, 25, 40, 60} except the artificial score of -1. A score of -1 means the pigs came to rest touching each another. -COLUMN 5 (height): The height from which the pigs were rolled. Values are the integers 1 = (5 inch) and 0 = (8 inch). -COLUMN 6 (start): The starting position of the rolled pigs. Values are the integers 0 = (Both pigs forward) 1 = (Both pigs backward) 2 = (One backward, one forward) SPECIAL NOTES: To keep track of the virtually indistinguishable pigs, we marked the snout of one pig with a black marker. If one pig lands in position 7, so must the other pig. This outcome ends the turn of the roller, and sets the total score for the roller to zero. Any outcome yielding a score of zero points ends the turn of the roller, and sets the score earned for that turn (and only that turn) to zero. PEDAGOGICAL NOTES: These data provide an opportunity for students of Bayesian inference to estimate the ten scoring probabilities using a multinomial-Dirichlet model. Of special interest is the availability of prior information: the accompanying scores allow for transparent prior specification (up to a single strength-of-prior-belief parameter). The Dirichlet posterior provides point estimates for the unknown scoring probabilities, and allows for simulation from the posterior predictive distribution of the next roll score. Independent realizations from this predictive distribution are then used to examine the effectiveness of various strategies. Classical analyses can be employed to test for the symmetry between the black an pink pigs, as well as for the effect of the roll height or starting position. SUBMITTED BY: John C. Kern II Duquesne University 600 Forbes Avenue 440 College Hall Pittsburgh, PA 15282 kern@mathcs.duq.edu