Table 3. Number of Draws.
State | Pr() | 95% CI | |
Pr(AbsClass = Abs) | 0.949 | [0.908, 0.990] | |
Arkansas | Pr(VotClass = Vot) | 0.957 | [0.913, 0.981] |
Pr(NMClass = NM) | 0.988 | ||
Pr(AbsClass = Abs) | 0.932 | [0.912, 0.980] | |
California | Pr(VotClass = Vot) | 0.951 | [0.919, 0.978] |
Pr(NMClass = NM) | 0.987 | ||
Pr(AbsClass = Abs) | 0.888 | [0.916, 0.978] | |
Connecticut | Pr(VotClass = Vot) | 0.988 | [0.912, 0.981] |
Pr(NMClass = NM) | 0.994 | ||
Pr(AbsClass = Abs) | 0.937 | [0.914, 0.977] | |
Florida | Pr(VotClass = Vot) | 0.970 | [0.917, 0.982] |
Pr(NMClass = NM) | 0.998 | ||
Pr(AbsClass = Abs) | 0.928 | [0.915, 0.980] | |
Kansas | Pr(VotClass = Vot) | 0.953 | [0.918, 0.982] |
Pr(NMClass = NM) | 0.996 | ||
Pr(AbsClass = Abs) | 0.907 | [0.921, 0.978] | |
Kentucky | Pr(VotClass = Vot) | 0.974 | [0.921, 0.978] |
Pr(NMClass = NM) | 0.993 | ||
Pr(AbsClass = Abs) | 0.979 | [0.915, 0.986] | |
Missouri | Pr(VotClass = Vot) | 0.947 | [0.904, 0.982] |
Pr(NMClass = NM) | 0.999 | ||
Pr(AbsClass = Abs) | 0.945 | [0.908, 0.991] | |
New Jersey | Pr(VotClass = Vot) | 1.000 | [0.895, 0.987] |
Pr(NMClass = NM) | 0.999 | ||
Pr(AbsClass = Abs) | 0.970 | [0.917, 0.982] | |
New York | Pr(VotClass = Vot) | 0.947 | [0.908, 0.985] |
Pr(NMClass = NM) | 0.987 | ||
Pr(AbsClass = Abs) | 0.941 | [0.911, 0.985] | |
Nevada | Pr(VotClass = Vot) | 0.963 | [0.915, 0.982] |
Pr(NMClass = NM) | 0.996 | ||
Pr(AbsClass = Abs) | 0.950 | [0.914, 0.986] | |
Oklahoma | Pr(VotClass = Vot) | 0.940 | [0.920, 0.980] |
Pr(NMClass = NM) | 0.998 | ||
Pr(AbsClass = Abs) | 0.975 | [0.912, 0.981] | |
Pennsylvania | Pr(VotClass = Vot) | 0.971 | [0.914, 0.986] |
Pr(NMClass = NM) | 0.994 | ||
Pr(AbsClass = Abs) | 0.972 | [0.908, 0.979] | |
Rhode Island | Pr(VotClass = Vot) | 0.953 | [0.912, 0.980] |
Pr(NMClass = NM) | 0.997 |
Yahtzee classifier results from 1000 randomly selected Facebook users from each state. Each user was given a classification based on the Yahtzee process: “Abs” Abstainer, “Vot” Voter, “NM” Not Matched. The conditional probabilities are calculated as the probability of observing a true behavior conditional on the Yahtzee classification. The 95% confidence intervals are for the null distribution of 95% accuracy in the classification, calculated from a binomial distribution with the same number of draws in each category. In total, 22 of the 26 tests fall within these intervals, suggesting that deviations from 95% accuracy are due to sampling variation, and for a large sample the procedure will generate the desired level of accuracy.