scary statistics: understanding risk for journalists

Upload: national-press-foundation

Post on 04-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    1/34

    An adventure in the perception of news

    Understanding Risk

    Rebecca Goldin, Ph.D.Director of Research, STATS

    Professor of Mathematics, GMU

    January 17, 2013

    National Press Foundation

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    2/34

    Statistical Assessment Servicewww.stats.orgJon Entine, Senior Fellow

    Cynthia Merrick, Intern

    Rebecca Goldin, Director of

    Research

    Trevor Butterworth, Editor

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    3/34

    Statistical Concepts in

    Writing about RiskMean, median, mode

    Standard deviation

    Confidence intervals

    Orders of magnitude

    Confounding factors

    Percentages

    Absolute vs. relative risk

    Scientific methods

    Causation versus correlation

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    4/34

    Absolute versus relative risk

    Absolute risk is the risk you actually undergo.Women who take the birth control pill have an

    absolute risk of venous thrombosis (blood clot) of

    about 1 in 10,000 per year. The absolute risk of

    women who do not take the pill is 1 in 15,000 peryear.

    Relative risk is a risk compared to another group.

    Women who take the birth control pill have a 50%

    increased risk of venous thrombosis, compared to

    women who dont take the pill.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    5/34

    Relative risk

    representations

    haveconsequence

    In 1995, Committee on

    Safety of Medicine in UKconcluded that the 3rd

    generation birth control pill

    was riskier than previous

    versions.

    Some press reported a

    100 percent increase in

    risk in Deep Vein

    Thrombosis (blood clots);

    others reported twice therisk.

    The absolute riskfor DVT was

    15 per 100,000 for 2nd

    generation birth control pills

    The absolute risk for DVT was30 per 100,000 for 3rd

    generation birth control pills.

    The media blitz led to many

    women not taking theirmedications (rather than

    immediately replacing them)

    and an increase in

    unwanted pregnanciesThe abortion rate went up 9%

    from 1995 to 1996.

    The absolute rate of DVT for

    pregnant women is 80 per

    100,000.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    6/34

    Kids dying from the flu

    Detroit Free Press (Jan 16, 2013):Risk to allages: About 100 children die of flu each year

    Story has many specific examples of children dying,with parents describing the tragedies

    Prevalence of flu vaccine discussed (About 40% ofchildren are vaccinated)

    Effectiveness of flu discussed (60% of people whowould have gotten it do not). Most deaths amongthose not vaccinated

    24,000 per year die of all ages in the U.S.

    No mention ofabsolute risk: there are about 74million children in the United States. The risk of deathby flu is 1 in 740,000 per year, or .00014 percent.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    7/34

    Comparisons are powerful for

    risk

    Cars Guns

    ~250,000,000registered cars in theUnited States(Department ofTransportation)

    ~32,000 car crashfatalities (NationalHighway Traffic Safety

    Administration, 2010) 12.3 road fatalities per

    100,000 people, 1.5-3times as many as inEurope (8.7 in Italy, 6.9

    in France, 4.5 inGermany)

    ~130,000 federallylicensed firearms dealersin the United States

    (Bureau of Alcohol,Tobacco, Firearms andExplosives); ~150,000gas stations, ~14,000McDonalds

    270,000,000 guns incirculation (survey, 2007))

    ~30,000 people die eachyear including 8,500 bymurder (2011, FBI),19,000 by suicide (CDC,

    2009), 1,000 by accident Rate of death by firearm

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    8/34

    Causation or Correlation

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    9/34

    Its easy to be fooled

    Height correlates with reading skills in childrenunder 10.

    Income correlates with success in college.

    Ratio of finger lengths correlates with aggression. Facebook correlates with poor grades.

    Facebook correlates with good grades.

    Doing heroin correlates with doing marijuana.

    Higher taxes correlate with high annual growth,and are inversely correlated with poverty rates.

    Alcoholism correlates with less gray matter in the

    prefrontal cortex.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    10/34

    IVF pregnancies may increase risk of blood

    clots, blocked arteries Fox News, Jan 16, 2013

    In vitro fertilization may come witha slight associated risk: blood clots

    and blockages.

    previous studies have found IVF

    to be just as safe as normalpregnancies, but [the authors of

    the new study] werent necessarily

    convinced

    Study Design:About 24,000

    women who had undergone IVFwere compared to about 115,000

    women who had normal

    pregnancies. Each had average

    age of 33.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    11/34

    Those who went IVF were more likely to haveblood clots. 4.2 out of 1000 had VenousThrombembolism among IVF women, comparedto 2.5 out of 1000 for other women.

    .08 percent of IVF women had a blocked artery,while only.05 percent of women with normal pregnanciesdid.

    But is it the IVF that increasedthe risk, or is IVF

    reflective of an increase level of risk? Women ineach group do not have the same health profile.

    Non-IVF contributors to these different numbers:different levels of fertility, access to medical careand diagnosis, hormonal treatments not directly

    related to IVF IVF re nancies carr more risk of blood clots

    IVF pregnancies may increase risk of blood

    clots, blocked arteries Fox News, Jan 16, 2013

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    12/34

    fMRI studies a case study fMRIs are large magnets

    measuring oxygen levels inblood

    People can engage in

    activities inside the machine Patterns of blood flow are

    thought to reflect patternsof brain activity (more on

    that in a bit). Typical studies: assume that observed patterns

    onlyoccur when the tested behavior occurs.

    Typical studies: assume that observed patternsare causedby the tested behavior.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    13/34

    fMRI studies a case study Lying can be determined by

    patterns of fMRI scans.

    But perhaps stress or anxietycan lead to the same patterns

    Violent video gaming leadsto violent brain patterns

    But perhaps any competitiveplay, including non-violent

    non-video games has similar brain patterns. Plus, noindication ofactualviolence.

    Math anxiety triggers activity in the pain center of thebrain.

    But no pain experienced by subject with math anxiety.Perha s anxiet , not mathematics, correlated with

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    14/34

    Jumping from Correlation to

    Cause You dont always have to know whyit may not be

    causal. Be wary ofanyclaims of causality.

    Some common reasons that a correlation could look

    causal when its not include: not adjusting for

    confounders, misunderstanding the mechanism,having an unknown confounder.

    A causal relationship might be reasonable to suspect

    when the statistics are

    Overwhelming Observed in many different contexts

    Repeated tests show the same effect, on large numbers

    of people

    Double blind case-control studies.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    15/34

    Causation vs. correlation is not the only

    thing to worry about in medical research

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    16/34

    The roll of randomness

    Given a hug urn of balls 30% of the balls are white,and the rest are other colors.

    Each of 100 people pick

    10 balls, write down their

    colors, then return the

    balls to the urn.

    Some people will have

    3 white balls, but others

    will have greater or fewer.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    17/34

    0

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0 1 2 3 4 5 6 7 8 9 10

    Probability

    Number of White Balls Chosen

    Randomness in Choosing 10 Balls:

    How many are white?

    Number of White Balls is

    Random

    Suppose that white represents something

    random, and bad, like the number of cancers per100 people in a town.

    If one town gets 7 cancers per 100 (double the

    expected number), wouldnt you think theres a

    reason? Our statistics suggest maybe not. But

    About 27% chance

    you will get 3 white

    balls; its much more

    likely youll get some

    other number

    About 1% chance of

    getting 7 white balls

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    18/34

    0

    0.05

    0.1

    0.15

    0.2

    0.25

    0.3

    0 1 2 3 4 5 6 7 8 9 10

    Probability

    Number of White Balls Chosen

    Randomness in Choosing 10 Balls:How many are white?

    Randomness has structure

    You can predict how likelythe data is, if you assumethe probability of white is 30%.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    19/34

    On p-values (how likely am I to see the

    data I see, if the data are random?)

    Suppose you are flipping a coinmany times, and you think this coin

    is biased, because you arent

    getting close to heads and

    tails. How can you quantify yoursuspicion?

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    20/34

    On p-values (how likely am I to see the

    data I see, if the data are random?)

    The p-valueis a measurement based on the data you haveseen: it answers the question: if the coin werefair, how likelywould I be to see the data I am seeing? In other words, ifyou had a fair coin, is it reasonable to see the proportion ofheads/tails, or is it very unlikely to see that?

    If you flip 1000 times, and you get 520 heads, there is justunder a 10% chance of getting this many heads (or more). Incontrast, if you had 550 heads in 1000 flips, the chance ofthis happening randomly is only about

    .1%., i.e. very unlikely if the coin were fair. The biomedical communit enerall acce ts =.05 5% as

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    21/34

    Multiple Testing, or How to

    Guarantee Results

    Once you have a standard, like p

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    22/34

    What happens in the lab:

    Experiments Galore...

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    23/34

    What the

    rest of the

    world sees

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    24/34

    Metaphors for Bad Statistical

    Methods

    Drunk looking for his keys under the lamp post

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    25/34

    Metaphors for Bad Statistical

    Methods

    Texas Sharpshooter Fallacy

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    26/34

    Depression linked to increased

    stroke mortality Heartwire

    Based on data from NHANES I EpidemiologicFollow-up Study (National Health and Nutrition

    Examination Survey)

    But if you just comb the data, you are likelytofind something juicy. This doesnt mean itsuntrue.

    We need to be savvier about challenging

    scientists that take advantage of randomness togenerate spurious results

    "Depression is not currently routinely screened

    for after a patient has had a stroke," the author

    of the study said in an interview. "We think it

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    27/34

    Causation, Correlation and Risk

    Correlation is not always clear

    Causation is often inferred

    Risks are often over-sold

    A j li t ibl

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    28/34

    Media Impact is

    Great

    In 1998, Andrew

    Wakefield published astudy on 12 childrenwhich was the basis forthe belief that Autism isa result of vaccinations.

    Press repeatedlyreported these results,even though thescientific community wasunable to reproduce theresults. The existence ofthis study gave greatervoice to other studies

    A journalist was responsiblefor an investigation of thescientific integrity ofWakefields work.

    After his autism study wasdiscredited, most mediacoverage about vaccinationsreports both sides of thestory about whether

    vaccines are safe or not. However, the medical

    community almostuniversally endorsesvaccine and believes that

    vaccines are safe. Pockets of measles and

    croup due to vaccinationrefusal or lack of herd effecthave been found in the U.S.and in the UK.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    29/34

    Basic Advice for a Journalist with

    Limited Time and Ideas

    Read as a skeptic at all times. Avoid most

    conclusions of causality. A lot can be understood by even a cursory read (

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    30/34

    Research is routinely plagued

    Research is plaguedWhat can a journalistdo?

    Low levels ofsignificance

    Multiple testing

    No acknowledgement ofrandomness in researchdesign

    Lack ofcontext/repeatedexperiments

    Scientists dont knowhow to talk to

    journalists.

    But if you are looking tofind one scientist willing

    Write about the levels ofsignificance, bias,caveats

    Ask the researchersabout multiple testing.Did they adjust forthem?

    Write about absolute

    risks. Look for a body of

    research rather thanone specific paper

    Cite your sources!

    DONT INDICATE

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    31/34

    Doting on Data

    Lessons to be Learned

    There is no certainty, due to random effects.

    Dont over-estimate the ability of poor data to

    give answers. Also, lots of data is unavailable.

    Risks need to be contextualized.

    Consensus is extremely important

    The world is complicated; many things interact

    with each other. The public voice is at leastas loud as the

    scientific voice.

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    32/34

    To Life!

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    33/34

  • 7/30/2019 Scary Statistics: Understanding Risk for Journalists

    34/34

    Thank you!