were not a part of the vision. >> host: when you look at king having a pointed view of thomas jefferson what did you feel? >> guest: we wanted looking him directly looking at jefferson but now he is looking toward reagan airport. >> host: what do you think conversation they would have if they had a conversation? >> guest: what he did say. he wrote wonderful words. let's live up to them. we justify our independence on the basis of the declaration that all people are created equal and endowed with rights. so now we have an obligation as a nation if we justify our independence, we have to live up to it.

>> your book "martin's dream" is a wonderful dream. thank you for the book. i did something my teachers told me never to do. i a dog year some pages. >> host: never a better complement to an author. >> host: good luck with the book. thank you for your joining me on afterwards.

[applause] >> if you have read naked economics you know, i have a tenuous connection to mass. these books were written in part out of desperation. i had a drink earlier and we were reflecting on how we struggle in the economics and statistics class and make economics was written almost by accident i called my agent was unsuccessfully trying to sell a book on the gambling industry. a good booktv written but i

said i have to teach the economic class to journalists. the textbook is inappropriate i cannot find anybody -- anything to see why they would have this and there was a long pause and she said he will write it and it will be called economics for poets and i will read it. that is how naked economics was born and those were scored away and bored to death we said let's go back and do statistics that is even more mathematically daunting on the other hand,, economics is the same the statistics will get what has happened over your lifetime, go back 15 years, if you gave

someone your credit card and it went into a bucket and filed but nobody had a digital record, a computing power was more expensive. fast forward now they will scan the book they will take your credit card data and the confluence of data, and can digitize that in sheep computing power we know more about people and that is easy to manipulate to study for a good. i will read for short sections. the first will revisit my uncomfortable relationship with math and talk about a probability event that is

older but gives the power of statistics. if you know, of the underlying math you can use it for good. one is a wake up should we be more aware with what is happening? and then the last is open questions i think statistics will in form but one example is what will football look-alike in tender 15 years with head trauma? statisticians and researchers looked at that question. but there will be interesting developments over the near-term. the introduction is called why i hated calculus but loved statistics. a two-story. i always had an uncomfortable relationship with math. do not like numbers for the sake of numbers. i am not impressed by fancy

formulas with no real-world application. i particularly dislike high school calculus for the simple reason no one ever bothered to tell me why i needed to learn there. what is the area beneath the parabola? who cares? number daughter goes through and i still don't care. one of my great moments within senior year of high school at the end of first sinister advanced placement calculus. i was working for the exam were prepared the nation have been and already accepted the college, so as they stared at the final exam questions they looked completely unfamiliar. not they had trouble answering them but i do not even recognize what they were asking. but to paraphrase john russell the usually knew what i didn't know. this exam looked more greek

than usual for i flipped through the pages than surrendered. i walked to the front of the classroom where my calculus teacher and originally it said we will call her carol miller because that is her name carol smith was prop during the exam. i said i do not recognize the stuff on this exam. we had a contentious relationship suffice it to say she did not like me either. i can now admit i use my limited powers and a student association president to live schedule allston assembly so her class would be canceled. we did have flowers delivered to mrs. smith from a secret admirer just so we could see what would happen. and yes i did stop doing any

homework cattle once i got into college. so i said the material to not look familiar she was not sympathetic. charles she said loudly, the of basing their rows of desks, if you had steadied, and the material would look a lot more familiar. it was a compelling point*. in a few minutes, a bright and, a better touch listed and then i and whispered a few things for issue whispered back in a truly extraordinary thing happened. class. i need your attention. it appears i have given you the second semester exams by mistake. you are far enough into the

test period the whole eggs and had to be aborted and rescheduled. i cannot fully describe my euphoria. [laughter] i was going to marry a wonderful woman with three children and published books and visited the tires of the still the day that might calculus teacher got her comeuppance was a top five moment. now that i nearly failed the makeup exam does not diminish from this experience. that is where i come from. there is a sad footnote that by the time i got to graduate school, i was feeling a little more confident. i was at the woodrow wilson school and they have resources to offer e caan and stats class's with or

without calculus. by the second semester i thought i am ready again. i took it with calculus and gotten beaten up and fortunately this class about calculus was taught by ben bernanke. [laughter] so my hubris is that i did not study courage economics by the current fed chair. that is where i come from. i have come to appreciate the enormous power of these tools but can you tell me why i need to know this? so now i will read a section from chapter five basic probability, do not by the extended warranty on the $99 printer. that you can figure out but

this makes me seem a little old but i hope you recall this. >> 1981 the brewing company spent $1.7 million for a shockingly bold and risky marketing campaign for the of brand shall its beer. do you remember? so the campaign did not work. at halftime of the super bowl the company broadcast the alliant taste test schlitz beer against mitt the low. boulder yet, they did not pick a random beer drinker's day picked 100 nickle loeb triggers the culmination of a campaign running throughout the playoffs. five live television taste

test each with 100 consumers of competing brands conducting the blind taste test and each was promoted aggressively just like the playoffs they would say watch schlitz beer vs budweiser live during the playoffs. do you remember? the marketing message was clear they think they like another brand would prefer schlitz beer but for super bowl live that halftime schlitz beer hired the former referee to oversee the test given that risky nature with live tv if they assume they produced a delicious beer. not the case. [laughter] they only needed a mediocre

beer and a solid grasp of statistics to know this ploy, said termite deny use lightly would almost certainly worked out in its favor. most beers in the category tasted about the same. in the '80s. all beer was bad. ironically that was the fact this advertising campaign exploited. the typical beer drinker cannot tell schlitz from miller oracle up or budweiser proposal of blind taste test is a coin flip. on average half will peck schlitz and half will pick the challenging beer. this alone would not make an effective advertising campaign. you cannot tell the difference so you might is will drink schlitz. [laughter] and it they would absolutely

positively not want to do that with its own loyal customers half would pick the competing beer. the looks bad when said beard drinker chooses the competitor which is exactly is what they were trying to do. but the genius of the campaign conducted the taste test who said they preferred another beer. if it is really just a coin flip then they will pick schlitz. half of all but wait -- budweiser drinkers it looks good halftime with a former nfl referee in uniform conducting the taste test. still, it is live television. even if the statistician's determine with loads of private chiles that the typical the globe drinker

will pick schlitz half of the time what if those turned out to be quirky? blind test is a coin toss but what if they chose with the low by chance? we lined up the same 100 men and flip a coin it is possible there would flip 85 or 90 heads. that is the bad book that would be a disaster for the schlitz brand and wasting $1.7 million for live coverage on television. statistics to the rescue. if there was a super hero i have six sigma may and. [laughter] but here they would come to the headquarters to unveil the details of the binomial experiment or called the upper new the trial. the key characteristic is a

fixed number of trials each with two possible outcomes and a the probability of success is the same for each trial. i define success as the taster picking schlitz. reassume all trials are independent and has no impact on any other. with only disinformation a statistical superhero king chow kit the probability of all outcomes for the 100 trials. 52 schlitz salacious 42 minutes alone. we could use a computer to do the same thing the chances of all when hundred baking mix low were, take a guess?

it is at the number one with 27 members after you're more likely to be hit by an asteroid watching the trial. picking give accumulative probability they clearly would assuage the fears. but assume schlitz would be happy if 40 at of 100 would pick schlitz impressive with the bond -- blind taste test profess to be in the globe drinkers it was that good of the beer test is like a flip of a coin that basic probability tells us a 90%

chance they would pick schlitz 86% 45 would this was not risky at all. anybody guess what happened? [laughter] halftime 1981 exactly 50% of the middle of drinkers chose schlitz. probability -- probability is a remarkable tool and many leading ears were indistinguishable from one another. [laughter] that should make you encourage about the power probability and statistics. now i will make you scared. this is the end of the book and the question who gets to

know what about you. last summer we hired a new babies -- babysitter i explained a family background, i am a professor, my wife is a teacher. she said nine no. i go gold you. i will simultaneously relieved and mildly alarmed how much of my life was on the shore internet search for our capacity to gather and analyze data with cheap computing power and the internet is unique we will need new rules for the new era. put the power of data into perspective with target this is told in "the new york times" magazine. targets tries to increase

profits by understanding customers. for the most part it is a good thing they hire statisticians to do the predicted analytics described earlier, the sales data to figure out who buys what and why. nothing is inherently bad but target they will carry things you want to buy. by drill down on the example that the statisticians figure out. but targets has shot being patterns with the relationship that could last for decades. women in there second

trimester get into the stores more often. a writer for "the new york times" magazine saw the predictive analytic team to attract private sherbert -- shoppers. target probably regrets this. [laughter] target has the baby shower registry that women register in advance of the birth. there already target shoppers and have told the store now not only are they pregnant but when they are due. the target will figure out other women who demonstrate the same shopping patterns are probably pregnant. then often switch to an scented lotions, vitamin supplements, extra big bags of cotton balls.

who knew? [laughter] target predicted analytics cruise identifieidentifie d 25 products together made possible what they described as a pregnancy predictions gore. the point* of the analysis to send pregnant women pregnancy related coupons to hook them as a long-term shopper. how good was the model? "new york times" magazine and reported a man from minneapolis who walked into a target store demanding to see the manager. he was irate because is high school daughter was related with pregnancy coupons. she is still in high school and you are sending her clothes for baby clothes and cribs? you encouraging her to be pregnant? the manager apologize profusely even called the father several days later to

apologize again. but this time he was less i rates and said it turns out there have been activities in my house i have not been aware of. she is due and august. the statisticians figure out his daughter was pregnant before he did. that is their business and also not their business. it can be little more interest of. for that reason some companies now mask how much they no. if you are a pregnant woman you may get coupons in the mail for cribs and diapers along with a discount on a riding lawn mower and a free pair of bowling sox.

to you, it seems fortuitous that stuff came in the mail but in fact, the company knows you cannot cut your lawn or bold but it covers the tracks so that what it knows is not so spooky. >> now let me finish with another question. i like this. there are good things about statistics in that thing is and where we watched it unfold in realtime. how can we identify to reward good teachers and schools? my wife is a public school math teacher. we need good schools and teachers so it follows logically we ought to reward

good teachers while firing bad teachers and bad schools. how do we do that? tuscaroras give us the objective measure yet we know some students will do better on standardized test for other reasons. the simple solution is to evaluate based on the progress the students make. what they know when they started, what did they know one year later? the difference is the value-added. we can use the statistics to take into account the demographic characteristics such as race, income and performance which is a measure of the base line aptitude. if they make gains then they

can be deemed highly effective to now have statistical quality with precision. how do these evaluations work in practice? in 2012 near city took the plunge to publish the ratings of all 18,000 public-school teachers of the value added assessment the measured gains of test scores with characteristics the "l.a. times" published similar data in 2010. in both cases the reaction has been loud and mixed. arnie duncan has generally been supportive to provide information where none existed previously. after the data was published he told the times silence is not an option the obama administration has provided

incentives for value-added indicators to promote teachers. proponents rightfully point* out they are a huge potential up geisha improvement that gives zero weight to any performance and the other hand they have large margins of error to deliver misleading results. they spent more than $100,000 on newspaper advertising campaign built around the headline this is no way to rate a teacher. others argue it has false position that will be abused by parents and public officials who do not understand their limitations of this kind of assessment. doug's digger warns data is

inherently no easy. results are often based on a single test taken on a single day by a single group of students. all kinds of factors can lead to random fluctuations like the broken air conditioning unit. that correlation going year-to-year is only about zero point* 35. on the a hand a correlation of year-to-year performance with major league baseball players is also zero point* 35 as the of batting average for hitters. they are useful in is one tool in the process but said david gets less noisy when they have more for one

teacher with different classrooms. in the case of the new york city they have been prepped on every use of the value added data and inherent limitations. the public did not get everything. as a result assessments are a definitive guide to the good and the bad teachers. we like the rankings. even with the data did not support precision. stater offers a final warning of a different sort. had better be certain of the outcomes here measuring from the given a standardized test ms match what we care about in the long run we just assume it is the end to care about the unique data from the air force academy

suggest, not surprisingly that the test scores that glimmer now may not be cold in the future. here is the importance of getting good data. the academy here is unique that it randomly assigns cadets to different sections of standardized core courses as introductory calculus. this eliminates any potential selection defect -- defect pork you cannot pick your class or professor so we assume they get students with serious aptitudes unlike most they can select in or out. they also use the same exam of every course those from

uc davis exploited the elegant arrangement to answer one of the most important questions which professors are most effective? and then doing a good job the professors with less experience and your degrees from fancy universities best students who typically do better clearly these young motivated struck -- instructors are from harvard. they must use the same notes they use to 1978 and probably think power point is the energy drink. [laughter] except they not know what that is either. of is the the data tells us we should fire them or let them retire gracefully.

hold on. don't fire anybody at. the air force academy study had a relevant finding was student performance over a longer horizon. in math and science those who have more experience and highly credentialed instructors come to do better with the mandatory courses than those who have less experience in the introductory courses. one logical interpretation is a less experienced instructors more likely to teach to the test that has impressive exam scores and happy students with the instructor evaluation. but deal crusty professors that we nearly fired one paragraph ago focus less on the exam and more on the concepts that matter most in the following courses and in

life after the air force academy. we need to evaluate teachers and professors to make sure we do our right. long-term policy challenge bruited in statistics is develop a system that rewards the real value added. this is a work in progress. we'll stop there will be happy -- happy to answer any questions. thank you. [applause] >> thank you for writing this book and also "naked economics." hope it sells as well as the other statistic book. >> this is a sequel. in the sense well as with my publisher that book was published by norton and he

said update this. that was written the 1950's and sold many copies but it is not coincidence. spirit there are least 8,000 obvious that will buy of copy. [laughter] i am interested in getting to the policymakers. i worked with the clinton white house with the first things they did was hand us a memo to the president by charles schulz we did not have a nice book like this to explain statistics. is a foundation to spend $20,000 buying 535 copies? >> i hope so. [laughter] >> then they would get the 20% discount. >> in addition what should be due to give policymakers a little more aware how misleading their numbers

are? they rely on them so much respect i am not sure everybody cares. the best thing we can do, when we teach, we have to move away from the mechanics. this is how you calculate this test made on the desk after 1985 it could do that for you but what you need to know is what you're doing. like the hand-held grenade launcher. we're better off to tell you where not to point* to it is then explain how works. so what makes statistics go off the rail? if you get bad data and put

it into a formula that will spell out the most wrong of conclusions. there is a more intuitive explanation, and where the formulas come from. to what we learn the things i have been taught i had to go back and that's it? i and stand up. but the first time around, not so much. there has to be a more intuitive approach and how it can go egregiously wrong. may be the gates foundation? and then to have one for the office and one for home. [laughter] >> thank you. >> thank you for the top. i would like to talk about tamoxifen.

heavily studied, then promoted, reduce your risk of breast cancer by half. i talked with the specialists in looking at statistics i asked is in.more accurate to say that it doesn't reduce there risk by half but reduces the risk from half of the women who take it? and how does a woman know if she is in the half that works? no answer. you are right. >> i have no knowledge. but the talk of clinical trials for pharmaceutical products because you end up where statistics can steer us wrong. the basic point* applied to pharmaceutical products that

is using the mean are the media and but if you look at the mean household income it goes up and a nice healthy way with explosive income growth in people are not getting any wealthier at all. so then you say always use the media which is the middle point* of the distribution but to all people sitting on a barstool each making $25,000 then put bill gates sits down the average income goes through the roof but the media does not at all. so that is the right one. there is a guy in the books who wrote a book the median is not the message. if there is a drug

successful for 40 percent of the people who take it, they live another 25 years, that mean life expectancy looks very low. this is her descriptive statistics people choose those matters but the of place where clinical trials come up there appears to be publication by is that if i did a study that said watching television eight hours per day does not making healthier when i get it published? they would say who thought that it did? of research is based on probability. a few of us will find by chance that those who sat on the couch got healthier.

so when they shop the study is the allied air. but someone will say utilities and not get heart disease? we have to publish. it is a serious problem if you don't look and all the studies conducted, the ones that are of most interest that show this works. they tried to do do a better. there is some sense of findings that there is a study of the study to suggest we may not be getting the terribly active view. that is the best i can do without a specific knowledge. but you have to be very

cautious with the data. your question was good. >> i notice you are a professor at dartmouth college. that was my first teaching job. i was a research instructor in the math department. >> of a great place to start the academic career. like these as was written in probability and i just wanted to make a few comments. you spend a lot of time in the book which is important to a show is teaching a business statistics course was is a difference between business that's and stats? it is one in which every

observation has a dollar sign in front of it. [laughter] you cannot show that in the basic course with the lady contadora them fairy but you can give the intuitive description of what the theorem says. here's my reward after doing that. student evaluation. professor rose in france, you talk too much and explain to much. cut out the small talk and give us the formula. so i tried to do what you would like that then here is another story. i will get to my question. [laughter] i feel like the ghost of hamlet's father.

to hear from a colleague a student wrote an evaluation of a professor he said he makes it awfully tough for the average student to get to and a day. [laughter] civic that maybe my student now. >> with great inflation it is worse >> i'm glad you talked about the modeling to draw your attention to a topic i am interested in because teachers that i know are fired right and left because of abuses statistics. here in washington there was the fifth grade teacher who was fired because of the value added model. persons did not perform as well as predicted. it turns out the scores of

the previous grade level, fourth grade, where the benefit from cheating. the staff erased the incorrect answers so the students came in with a higher predictive value. she was worried because she saw their scores and they did not correlate with the classroom. there is of vulnerability that is flawed with cheating. i talk to the executive director of the american statistical association and he said there is a special committee that is analyzing the value added methodology.

by the way i just wrote a very simple paper i have done submitted for publication that has a simple binomial model that explains why a teacher can be rated very high one year below the next because of the natural variation. >> okay. those are terrific points. that will be with us. a couple of observations, one of the parts of the book is using probability to catch cheating on standardized tests. there are private firms that if you run the answer sheets through the scandal:they flag cheating. for example, a high proportion of the racers going from wrong to right it tells the story that in the

land to the huge sheeting scandal that the probability of observing those patterns was a probability as delaying the astrodome with people who were over 7 feet tall is very unlikely. then discover anything but to your point* what they discovered is teachers and administrators were having pizza parties on the weekend to they all changed the answer is that the same time. if we have to do this three you might as well do it the old same time. wife is a teacher at super high-stakes testing us especially charter schools they did not quite cheap to but you don't have to take anything down in your room

so my wife went to a teacher development day that was maximizing the surface area of your room including the door so your students could work looks somewhere to see the answers. that is what they studied. the was probably not the intent. last, the others said we like to take this test because they get chocolate. they cannot have a lot say. >> have you heard of campbell's lot? >> the greater is the social and economic consequences associated with a statistics such as test scores, the more likely it is the

statistic itself will become corrupted and its use will corrupt the social process it was intended to monitor simic this is great for i am writing a short piece for "the wall street journal." one headline is smart managers use statistics to evaluate employees, employees figure how to manipulate the statistics. in new york state there is one they decided they would provide data on mortality rates for angioplasty for cardiologists. but it turns out somebody followed up and a high proportion but 80 percent of cardiologists said they deliberately change behavior because of the evaluation. it was not to kill fewer

people like quote today i will not drink. [laughter] but on the margin they decided not to take the most seriously ill patients. that is one of the great takeaways. people who were the most in need of surgical intervention were released likely to get it. >> the most dangerous place in the world is the hospital >> that will be in the next edition. thank you. >> i feel silly but therein is where data mining and machine learning doing away with cognitive method. machine your way to the cure for cancer.

but then we cannot take the people out of the statistics. starting with a hypophysis with a theory based approach. i have seen reasonable arguments on both sides. what do you think? >> it is not either/or. i thought about using high-powered computers for medical diagnostics. there is value there. but my caution is the distinction between precision and accuracy. any statistical model king give a false sense of precision. and one funny example and one serious, my wife gave me the golf range finders for christmas. you look at the pine trees

and think i am 150 yards. the with their range finder you are 148 point* seven. it is amazing. but my golf kept getting worse. i was hitting the trees, another time of the municipal parking lot the then read the instructions it was set to leaders. [laughter] not yards. it was meters. but i was not looking around. the faulty speedometer is worse than no -- no speedometer. in this fall to use tear at it with snow speedometer you have no choice to look around at other cars but the serious example this may in the end answer your question

the value added risk model hovels street probability that would put all the rest of the firm into a single dollar figure. 99 out of 100 times the firm would not lose more than this some over the next 24 hours or any time period. these models assumed housing prices would not fall. so people were building fancy models and the underpinnings are rotten and it spits out meters that is dangerous. said the value added with medical outcomes are a tool but if you get too far away from human judgment or let the false precision drill your sense of what is

accurate, it is quite dangerous. but that is a great question and one we are wrestling with for a long time. >> can you address political chicanery? with the obamacare and payne and very successful but there is the bombastic comment under obama your 10 times more likely to be on food stamps to skew your thinking away from the wal-mart policy. >> there is not and express chapter tuesday's in politics. that would be a good chapter. instead, every of a chapter

the way the book is laid out here is the power of probability here is the abuse. and every one of those is something that will tend to show up in politics. i use economic data with the presidential campaign tries to take as economic indicators. there is a whole chapter on polling and discussion if you have bad data, it does not matter how big your sample is. it does not show up explicitly but one polling piece that was a surprise, a frank newport editor in chief of gallup came to

dartmouth and one irony of polling right now that unlike the other things, it is getting harder to conduct a sample that reflects the general population. it is cheaper than ever to gather the data. but think how you do it. with a phone sample to reduce a random set of the area codes but now all of my students have of sulfone it is 213, 405 and they live in new hampshire. you have lost track where they lived. rich people with call waiting who will not answer calls from strangers. lonely people will. [laughter] and young people without a land line at all.

so now we have to move away from the assumption ha one landline per household so now they go back to the methodology they started with which was in person going back out door-to-door like in the '50s. the general takeaway is beware of large samples. they are expensive and unless they are dropping many is click here if you think we are doing a good job. [laughter] stomach but the topic of your books what jumped out what is the finance company and the baby spokesperson says that is about the same chance as getting mauled by a polar bear and a regular bear in the same day. [laughter] >> the story came out after the book of a guy was struck

by lightning on his way to buy a lottery ticket. [laughter] i thought he had the one probability but not the one he was looking for. thank you for coming. [applause] >> we are here with the author of snowstorm in august. what happened in washington d.c. 8035? >> francis scott key, all the authorities lost control of the city. the beginning of the conflict over slavery and the white population would attack the freed black population that was active in the antislavery movement. widespread disorder.

he was responsible and it was a great humiliation for him and the city. who is arthur bowen? >> he was a servant 90 years old in the home of an emory s. horton, a well-known woman. he was alleged to have attacked her in her bedroom at night with an ax. it was a sensational news story that set off the white population to attack the black population but there was no attack. >> host: how did they find that out to? >> guest: arthur went on trial and washington was scandal -- was shocked when she said i know this boy he grew up in my house and did not intend to harm me. he was drunk. he was convicted anyway but

because of her persistent desire to free him, she managed to get a pardon from president jackson and was saved from it execution. >> host: what was her connection to the president? >> guest: she was the widow of a manually important -- soaring who designed the u.s. capitol was a friend of washington and jefferson. she was friends with dolly madison and james madison. she had access to the white house and she used all of her connections to rescue this boy from death row. >> with a very particular title snowstorm in august. what is the snowstorm? >> the name local people gave to the riot after word. one of the targets was a man named beverly's know who was a man who