You are on page 1of 6

Statistical Literacy II

Six Types of Bad Statistics . . .
I. Missing Numbers
II. Confusing Numbers
III. Scaryy Numbers
IV. Authoritative Numbers
V. Magical Numbers
VI. Contentious Numbers

I. Missing Numbers

Dan Rather claimed that “school shootings in this country have
become an epidemic” when a student in Santee, CA shot and killed
two other students and wounded 13 others (03/05/01).
 Were school shootings really a widespread and growing
phenomenon?
 There had been four from 1997 to 1999, and researchers
identified 321 violent deaths from 1992 to 2001, but not all of the
latter were student-on-student and even if we count all as such
we see that the number actually fell from 48 to 32 per year
 Also, on average, there are fewer than one in a million violent
deaths per year, suggesting a low frequency in decline

In sum, this missing information distorts and misinforms the public

1

CO sometimes overwhelm readily available subtle statistical evidence.000 and another who is opposed that more than half of all families will experience a reduction of less than $100?  A percentage is also one of our most useful types of statistics (e.Reasons for Missing Numbers • Atypical dramatic examples such as the 1999 Columbine High School shootings in Littleton.... police shootings of civilians). 10% of all adults . Census Bureau or other organizations (e.. although special purpose surveys may sometimes count these (e. movie theater shootings) • Still other statistics are sometimes deliberately not counted by the U. .g. ). religious affiliation. kidnapped children. 2007 BM-IHLSS) • And. depression. Confusing Numbers • Even simple numbers can be confusing  “Around 40% of Republicans will always vote for a Republican presidential candidate and about the same percentage of Democrats will vote for their party’s candidate.g. some numbers providing idi context are forgotten (e. skin color. .g. 1994. unauthorized migrants. major causes of death.g. but familiarity does not always translate into understanding. slaves. 2 . McKinlay & McKinlay 1977) Reasons for Missing Numbers II.S.. The battle is for the middle 20%” (Thomas WNJ 2003)?  How can one supportive politician claim that a federal tax cut will result in the average family’s tax bill going down $1.g. • Other times statistics are not readily available because they are hard to count (e. 2001 & 2012 LAC-MIHLSS.

Confusing Numbers  The role of intentional selectivity (manipulation) 3 . Confusing Numbers • Simple graphs and other pictures can also be confusing  The role of aesthetics II.Delinquency and Sinistrality II.

etc.. . Scary Numbers  Why are risk estimates rarely generated from experiments involving human subjects (Best 2004: 79)?  Statistical theory.g. frail bones) threefold (or by 200%) to warrant attention (Best 2004: 81) . drinking soda) must increase the probability of an outcome (e. which is 200% of 5 .. among 1.e. .. if P[Lefty]=0. .III.But what if 60 in 300 were “frail” and soda drinkers were 20% more likely? We would expect to find that 72 of 300 soda drinkers are frail . iatrogenesis – Best 2004: 67)  Changing numbers (e.90 . then P[3 Lefties in a row]=0. . i Y2K Y2K.05% prevalence) have weak bones and soda drinkers are 20% more likely.g.each sequence (HHHHHH or HTHTHT) is equally likely  Thus. . . . global l b lh heating) i ) • Risks often are associated with everyday activities such as eating. things appearing to be abnormal may simply be rare . be suspicious of reversing trends and explanations based on anxiety linked to fast-changing society)  Apocalyptic A l i numbers b (M (Malthusian l h i starvation.07) women are “false positives”  Only 9% of “positives” (7 of all 76 positives) actually have it! III.000 soda drinkers would be frail . Prob[Woman in 40s Has Breast Cancer] = 0. and be aware of “probability neglect” (Best 2004: 83) 4 . Scary Numbers  What is the probability that a woman in her 40s who receives a positive mammogram test result (say.000) has breast cancer if . and involve “probability” (not well understood)  What are the odds of “heads” after flipping four in a row? . . .g. then three times this is 15.10.90) women are correctly diagnosed Prob[Incorrect Positive Mammogram Diagnosis] = 0.. .g.000 (.about 69 (992 x 0.If X=5.. .8 women have breast cancer . Scary Numbers • Everyday we are bombarded with scary numbers that are meant to generate fear and are often presented out of context.  Big numbers (e. Is this weak evidence?  We should also insist on multiple studies rather than relying on one . then 6 (rather than 15) in 10.992 do not have breast cancer Prob[Correct Positive Mammogram Diagnosis] = 0.008 .001 III. . suggests a factor (e. .07 ..about 7 (8 x 0.. sleeping. .If 5 in 10. Is this strong or weak evidence? .g.

. crime rates.However.g. and divorce rates need to be computedd as a fraction f i off the h population l i off interest i . accurate. .S. News & World Report’s “swimsuit issue” . and that promise to make our problem more manageable. interestingly this is ≈ 50% IV.. Magical Numbers • Magical numbers are those we imagine to be authoritative.Typically computed by dividing the number of divorces by the number of marriages in a given year.Thus. Were black teenage “suicides” suicides really rising in the late 1990s as the CDC reported?  How many people were killed on 9/11/01 (6.III.College admission process (e. Authoritative Numbers • Authoritative numbers are those produced or presented by those thought to be authorities.  Selecting a college and use of U. SEC.Death rates.External (e.  Even the most professionally compiled data (e.Do standardized tests measure more than socioeconomic class?  Racial Profiling by Police .g.So. 1945-1996) by the number which ended in divorce . ..Characteristics of accepted students (e. many suggest the divorce rate is quite low and marriage is quite a stable institution (e.  Recall the JAMA “bullying” example we discussed in week one. What is a good measure of school quality? . this would be the result if 50% of all marriages each year ended in divorce because the base keeps growing . what we need to do is divide all marriages that occurred during a long period of time (e.. divorces in 2011/all married people) .g. (Best 2004: 94-103).Peer ((other colleges  Business and other Institutional Decision-Making .Should we compare the percentage of those stopped to that of those who reside somewhere or of licensed drivers by race? 5 .g.g.. but this is not how a rate is computed! . where it was reported that 33% of students in grades 6-10 have been bullied or bullied others .500 or about 3. .. death records) can be misinterpreted . earnings. .g. SAT) . .Since 1960 this has been about 50% . . admission rate) g & universities)) assessments – 40% . . Scary Numbers  Risk of Divorce = 50%? .Internal (bureaucratic) numbers games . FBI) numerical manipulations  School Quality and Import ..000)? • But what is often thought to be authoritative is misleading V.

. Contentious Numbers • Contentious numbers may involve missing.VI. scary.g. but they employ or interpret them ideologically (e. h equally ll valid. confusing. and we ought to desire more and better evidence continually • Debates rarely start over a number. authoritative or magical numbers but are used to win an argument  Best (2004) argues that we ought to try to avoid both relativism and extremism when considering numbers  We should also not accept the notion that a statistical view is just one among many other. child obesity due to food industry or federal government’s school lunch programs?) or selectively • Estimates of the Numbers of Jews and Muslims in the USA 6 . lid ones  Single studies or numbers are not sufficient.