V1a 4/29/2017 Statistical Literacy: 2017 V1A 2017 CME 1 V1A 2017 CME 2 . Audience . 40%: School teachers: Current or previous 30%: College faculty: Current or previous 20%: Education, non-profit 10%: Industry, commercial V1A 2017 CME 3 V1A 2017 CME 4 Statistical Literacy 2017: Statistical Literacy 2017 Overview “We teach the wrong stuff ; We teach it the wrong way ; Milo Schield, Augsburg College We teach it in the wrong order .” Richard de Veaux Elected Member: International Statistical Institute US Rep: International Statistical Literacy Project Statistical Literacy 2017: VP. National Numeracy Network 1. What is it – in general? CME Presentation in Toronto 2. Who needs it? Fields Institute 3. What is it – in particular? April 29, 2016 4. Who can implement it? www.StatLit.org/pdf/2017-Schield-CME-Slides.pdf 1a 1a V1A 201 CME 5 V1A 201 CME 6 Statistics is Different What are Statistics? from Mathematics a. Data; numerical data, classifications of data, Math ignores the context. or numerical summaries of data [Ambiguous] a. Math deals with form (ignores the matter) b. Math deals with variables and values (no natures) b. Outcomes from a random process; c. Math deals with associations and co-variates randomly-selected or randomly-assigned groups d. Math has no operator for “causes” [Technical distinction] Statistics depends on the context a. Statistics deals with the matter: its nature c. Numbers in context where the context matters : b. Statistics deals with subjects and characteristics Quantitative summaries of real things: things that have natures, connections & causes c. Statistics deals with “confounders” d. Statistics deals with “causes” www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 1
V1a 4/29/2017 Statistical Literacy: 2017 1a 1a V1A 201 CME 7 V1A Saying 2017 CME Mathematics: “Statistics Come From Data” Patterns vs. Nature is like saying . “Babies Come from Hospitals” It’s true but it leaves out the interesting details Statistics are answers to questions or interests. 1b 1b 9 10 V1A 201 CME V1A 201 CME What is statistical literacy? What is statistical literacy? In general terms Examples: Statistical literacy is needed by citizens and social Medical: decision makers to enable them to understand and • Describe: Japanese, who live long, eat low-fat diet evaluate the statistics they encounter everyday. • Compare: People with high-fat diets die sooner Everyday statistics are used as evidence in arguments. • Evaluate: High-fat diet causes shorter lifespan. Legal: Social: • Describe: 90% of a restaurant’s staff speaks Spanish • Describe: Average school class size is 24 • Compare: Most Mexican restaurant staff speak Spanish • Compare: Best performing classes are smaller • Evaluate: Mexican restaurants discriminate in hiring . • Evaluate: Smaller classes will improve outcomes 2 2 12 V1A 201 CME 11 V1A 201 CME Three Audiences: Who Needs Statistical Literacy? More detail Three Audiences . 1. STEM majors and those who conduct surveys, studies and research. 2. Social decision-makers: Politicians, bureaucrats, business leaders, doctors • Those who inform citizens and decision makers: journalists, analysts, lawyers, economists, consultants, sociologists, political scientists, policy advocates, psychologists and educators. 3. Citizen in a modern republic or democracy. www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 2
V1a 4/29/2017 Statistical Literacy: 2017 2a 2a V1A 201 CME 13 V1A 201 CME 14 College-Bound Students: College-Bound US Students Wide variation SAT Math Scores by Major . SAT (CR+M): US College-Bound Seniors 1600 Top 25 Colleges 1400 St. Thomas 1203 Augsburg 1200 1070 1000 Community 800 Colleges Mean: 1010 600 StdDev: 218 400 0 20 40 60 80 100 CollegeBoard Percentile Business Insider (2014). 2014 SAT scores 2014 2a 2 V1A 201 CME 15 V1A 201 CME 16 US Stat 101 students Harvard Business Review: by Major Website Search of 40K Items • . 3 3a V1A 201 CME 17 V1A 2017 CME Statistical Literacy: Association: Probably Not Causation More Detail 3a. Association vs. Causation . 3b. Classify all the influences on a statistic • Context: Confounding and study design • Assembly/assumptions: How things are defined. • Randomness: Unlikely is almost certain given enough trials. • Error/Bias: www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 3
V1a 4/29/2017 Statistical Literacy: 2017 3a 3a V1A 2017 CME V1A 2017 CME Association: Probably Causation Association: Possibly Causation Heart-Attack Survival Rate . 40% 30% Survive 20% 10% 0% 0 1 2 3 4 5 6 7 8 Minutes to Defibrillation Jolt delayed in third of cardiac arrests AP Story: 01/03/2008 3a 3a 21 22 V1A 2017 CME V1A 201 CME Distinguish Causation Action-Verb Association from Association C ausation (8%): cause, effects, results, prevents “Research shows that the headgear A ssociation (2%): associate, relate, correlate, reduces the B etween (67%): concussion rate Action verbs: ups, cuts, raises, boosts, increases by more than Other: due to, because of, attributed to 50 percent.” Inappropriate use of “causes”: • Obesity causes later onset of puberty in boys 8/2011 P. 41 • Junk food causes a third of heart attacks. Schield and Raymond (2009) study 2,000 newspaper headlines involving quantity 3a 3a V1A 201 CME 23 V1A 201 CME 24 Association-Causation Pie Chart: Compare Baseball players whose names begin with the letter Protestants are twice as likely to be smokers as are “D” are more likely to die young Catholics ? SMOKERS Drinking a full pot of coffee every morning will add years to your life, but one cup a day increases Catholics: NO: the risk of pancreatic cancer. 20% Smoker is Other: 40% whole. Asian-Americans are most susceptible to heart attacks on the fourth day of the month Student error rate: Source: Standard Deviations: Flawed Assumptions, Tortured Data, Protestants: 40% 62% and Other Ways to Lie with Statistics by Gary Smith (2015). www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 4
V1a 4/29/2017 Statistical Literacy: 2017 26 V1A 201 CME 25 V1A 2017 CME Association vs. Causation Speculative [Spotty] Statistics Using Ordinary English 11 Headlines, Same Story Does a death certificate ever list air pollution 1. Study: 45,000 Uninsured Die a Year (CBS News) as a cause of death? Does a coroner certify this? 2. 45,000 deaths attributable to uninsurance These are association-based statistics. 3. 45,000 US deaths associated with lack of insurance 4. No health coverage tied to 45,000 deaths a year These are speculative 5. Lack of insurance linked to 45,000 deaths (spotty) statistics. 6. Study: 45,000 U.S. Deaths From Lack of Insurance 7. One death every 12 minutes due to no health insurance 8. 45,000 ... die because of lack of health insurance 9. Lack of Health Insurance Kills 45,000 a Year 10. Lack of Health Insurance cause 44,789 deaths 11. Lack of insurance to blame for almost 45,000 deaths 3b V1A 201 CME 27 V1A 201 CME 28 Stats = Premise: Crit. Thinking Statistical Literacy in detail: Stats = Conclusion: Stat Literacy “Take CARE” . Statistical literacy studies all influences on statistic: • Confounding: • what was – and was not – controlled for • what kind of study was involved • Assembly/Assumptions: • how statistics are collected, defined and grouped • how statistics are summarized, compared & presented • Randomness: small samples and big data • Error/bias 3b 30 V1A 201 CME 29 V1A 2017 CME Small Change in Syntax; Confounding: Using Ordinary English Big Change in Semantics 1) The percentage of women who are runners. . 2) The percentage of women among runners. 3) The death rate of men is X per 100,000. 4) The men’s rate of death is X per 100,000 5) Toyota is the car most frequently stolen. 6) Toyota is the car most likely to be stolen. 7) Cadillac is the car most likely to be stolen. Edison 2009/09/26 www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 5
V1a 4/29/2017 Statistical Literacy: 2017 32 V1A 201 CME 31 V1A 2017 CME SEASON WINS vs. TOTAL PAYROLL Confounding: Mixed-Fruit vs. Size of a statistic depends US Major League Baseball Apples-Apples Comparison 102 on what is “taken into account” Indians . 92 Braves 1995 Season Wins Red Sox Reds 82 27# Yankees Rangers 72 Mets Padres Orioles Marlins Expos 62 Tigers Pirates 13# Twins BlueJays 14# 52 10 20 30 40 50 60 Total Payroll ($Millions) 33 34 V1A 2017 CME V1A 2017 CME Study design can inhibit US SAT-VERBAL SCORES certain kinds of confounders . 3a 36 V1A 201 CME 35 V1A 2017 CME A ssembly: Assembly: What to control for: United has Worst Pet Record Making small things big 7 nanograms per gram = 7 parts in a billion Nine pets died while being transported by United while another 14 were injured last year. Most of any US airline… 4/2010 National Geographic www.StatLit.org/pdf/2017-Schield-CME=Slides.pdf Page 6
Recommend
More recommend
