Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Classroom Simulation: The Margin of Error in a Public Opinion Poll JSM 2005 — Minneapolis Bruce E. Trumbo Eric A. Suess Shuhei Okumura California State University, East Bay (formerly CSU Hayward) [email protected] [email protected] Pedagogy of Probability Euclid formulated axioms of geometry 2300 years ago. Kolmogorov's axioms of probability: 1935 Why the time gap? Probability is harder. “Practical experience” precedes formalism. You can draw a triangle with a stick in the sand. Pyramid builders used geometry. Probability revealed in “large numbers.” “Labs”: Gambling casinos began in 1600s. Role of Simulation in Class Small-scale random experiments using real objects can be used for introduction. (Coins, dice, drawing beads from a bag.) But time consuming, yield small-scale results. Computer simulation provides large-scale results quickly. High-quality software necessary. Minitab, SPSS, SAS, S-Plus, R are OK (not Excel). R excellent for simulations and free, but not point-and-click: www.r-project.org. Presentations at Various Levels Basic statistics “service” course: Present as a slide show. No discussion of R programming. Prelude to discussion of normal (and possibly normal approximation to binomial). Calculus prereq. statistics/probability course: Run multiple simulations in R. Discuss normal and binomial distributions. Perhaps introduce simple R code. Senior or first-year MS level: Also discuss Feller’s Arcsine Law. Illustration: “Margin of Error” in a Public Opinion Poll Many polls quote a margin of sampling error based on the number n of people sampled: 2% for n = 2500 3% for n = 1100 4% for n = 625 For a candidate's “lead”: Double the margin of error. What is the basis for these numbers? Assume a random sample. Ignore administrative / subjective issues (nonresponse, phrasing of questions, deceptive answers, etc.). Margin of error based on probability rules: hard to prove at an elementary level. Simulation illustrates these rules. Experimental, exploratory approach. Simulating an Election Poll To simplify, suppose: No undecided voters in population. 53% of all voters favor Candidate “A”. 47% of all voters favor Candidate “B”. Randomly sample 25 voters. Can we detect from data on 25 subjects that Candidate “A” is in the lead? One Simulation Using R > sample(0:1, 25, rep=T, prob=c(.47, .53)) [1] 1 0 1 1 0 0 1 0 0 1 0 0 0 [14] 0 1 0 1 0 1 1 1 0 1 1 0 Interpret 0s & 1s as candidate preferences A B A A B B A B B A B B B B A B A B A A A B A A B Only 12 / 25 = 48% for “A”. Misleading result. What can we conclude from this small poll? Endpoint is 48%. What would other polls say? Note: Here and beyond, vertical axis from 25% to 75% to help show detail. Results Differ Widely. Next: On one graph, overlay 20 traces. All 20 traces end between 40% and 70%, most above 50%. Of 1000 traces, 377 (more than 1/3) end below 50%. These give a false indication that P(A) may be less than 1/2. A 25-subject poll can’t reliably show if “A” is winning. Now Consider Polls with n = 2500 Subjects Major polling organizations report such polls as having a 2% margin of sampling error. Intended Meaning: If the true proportion for “A” is 53%, then most 2500-subject polls (95%) will have endpoints between 51% and 55%, thus detecting that “A” is the favorite. Short marks at right indicate 53% 2%. Reference line at 50%. On “average” run: 19 out of 20 traces (95%) within margin of error. 95% of endpoints within 2% margin of error; only 3 in 1000 below 50%. In the interval from 51% to 55% for Candidate “A” the peaked curve has 95% probability; the flat curve only about 12%. Feller's Arcsine Law In above demos of margin of error and normal distribution: P(A) = 53% and we focus on the endpoint of each trace. Now change focus: Now P(A) = P(Heads) = 1/2 (think of a fair coin). For each of 10,000 traces, find the percentage of steps where the trace is above 1/2. Make a histogram of these percentages. Many students expect results to cluster around 1/2. But actual cluster points are near 0 or 1. Histogram approximates BETA(1/2, 1/2), which has an “arcsine” function as its CDF. 0.0 0.5 1.0 1.5 2.0 5000 Experiments Each With 10000 Tosses of a Fair Coin 0.0 0.2 0.4 0.6 0.8 1.0 Proportion of Time Head Count Exceeds Tail Count Based on W. Feller (1957): Intro. to Probability Theory and Its Applications, Vol. I [The "bathtub shaped" distribution is BETA(1/2, 1/2).] In any one trace, either Heads or Tails is likely to stay fairly consistently in the lead. References Moore, David S.: Basic Practice of Statistics, 3rd ed., W. H. Freeman (2004) p479. www.sci.csueastbay.edu/statistics/ Resources/Quiz/poll.htm Feller, William: An intro. to probability theory and its applications, Vol. 1, 3 rd ed., Wiley (1968) Trumbo, Bruce E.: RU Simulating Column: “Polls on steroids,” STATS magazine, Fall 2005. (Similar to parts of this poster presentation.) www.r-project.org (Instructions for downloading and installing R.) www.sci.csueastbay.edu/~btrumbo/JSM/pollsim (R code, additional materials for this paper.) Other S-Language Classroom Simulations at a Similar Level. Trumbo, B. E.; Suess, E. A.; Schupp, C. W.: “Using R to compute probabilities of matching birthdays,” Proc. JSM 2004. (Also see: STATS, Spring 2005.) Trumbo, B. E., Suess, E. A; Fraser, C. M.: “Using Computer Simulation to Investigate Relationships Between the Sample Mean and Standard Deviation,” STATS, Fall 2001, pages 10-15. Trumbo, B. E., Suess, E. A; Fraser, C. M.: “Contemporary statistical simulation methodology for undergraduates,” Proc. JSM 2000. (Also see: STATS, Winter 2000.)