Stephen T. Ziliak is Professor of Economics at Roosevelt University and Conjoint Professor of Business and Law at the University of Newcastle-Australia. A major contributor to the American Statistical Association “Statement on Statistical Significance and P-values” (2016) he is probably best known for his book (with Deirdre N. McCloskey) on The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice, and Lives (2008), showing the damage done by a culture of mindless significance testing, the history of wrong turns, and the benefits which could be enjoyed by returning to Bayesian and Guinnessometric roots.
+ Full Transcript
Rosemary Pennington: Back in September, we asked you to explain Bayesian analysis to us as simply as you could in a journalistic headline and lead sentence. Forty of you submitted entries using everything from Netflix to the moon to newspapers to help you explain Bayes. Today on Stats and Stories, where we explore the statistics behind the stories and the stories behind the statistics, we talk with our Better Bayes winner! Stats and stories is a production of Miami University's departments of Statistics and Media Journalism and Film as well as the American Statistical Association. I'm Rosemary Pennington. Joining me in the studio is regular panelist John Bailer chair of Miami's Statistics department. Media journalism and Film’s Richard Campbell is away today. Before we tell you who won the better Bayes competition, John and I are going to share the four runners up.
John Bailer: We should have a drumroll right now!
Pennington: I know! I’m not talented enough to do that. I should have thought about that before we came in here. So I'm going to read the first was Lucy D'Agostino McGowan, her headline was “Using Bayes to binge, A Netflix Original Series” and her lead deciding which Netflix series to binge based on prior shows you've enjoyed pick one. One episode in, decide if you like it. Not sure? Watch another episode. You’re a Bayesian.
Bailer: And the second in the list of runner ups is Rasmus Bååth, with the headline, “Bayes’ uncertainty modeling”. Use the language of probability to describe what you know and what is uncertain about a situation. Add data, voila! Bayes uses the new information to optimally reduce the uncertainty.
Pennington: Our 3rd runner up, not in 1, 2, 3 order just on the list of runners up is Alex Holcombe. His headline is “A Bayesian film noir”, with the lead round up the usual subsects, Sargent! But sir, we haven't a clue! True, we don't but we do have a prior, later in the story. Help me update my prior, fryer! Rev bays didn't laugh but he got to work.
Bailer: And the last on our list of runner ups is Joshua Bon’s contribution, with headline, “How (un)certain are you” and that was, the un was parenthetically, certain or uncertain are you, with a lead sentence, imagine choosing your route to work but without Google Maps. Your decision will depend on some data, for example, the weather and your prior experience driving these routes. Bayesian analysis is a mathematically principled way to combine data and prior information for decision making under uncertainty.
Pennington: About 40 of you, as we said, entered the contest. Our winner is actually a professor of economics at Roosevelt University, Steve Ziliak, and we have him here with us. Thank you so much for being here today, Steve.
Steve Ziliak: Thank you so much for having me. I'm delighted!
Pennington: Do you have your entry handy?
Ziliak: I do.
Pennington: Would you mind reading it for us?
Ziliak: Not at all! It will be my pleasure.
Better Bayes, found in linked haiku.
Bayesian method- Making stuff you partly know Link that with what you don’t.
Frequentist methods Fear the null hypothesis and large p-values
“It pays to go Bayes” Epistemologically Obvious in prior.
Pennington: It’s great. I want to know John, do you fear the null hypothesis?
Bailer: Perhaps I'm just not a dyed in the wool frequentist. I'm more pragmatic in my particular belief system. No, I thought this was great.
Pennington: Yeah. So unfortunately our judges who were Kerrie Mengersen, who was a distinguished professor at Queensland University of Technology and Annie Laurie Blair who is a professor of journalism here at Miami were unable to join us and we were going to ask them why they chose a haiku instead of a news lead so since we don't have them I'm going to ask you why you decided to submit a haiku to the contest and how you came up with this?
Bailer: You took my question, I wanted to ask that one!
Pennington: You got to be faster, John!
Ziliak: Thank you so much again. When I saw it, I actually saw the advertisement for the competition on Twitter and I just got so excited. I just cracked up laughing as soon as I saw it and actually it didn't take me long to write my entry I felt so inspired, I really did, by the constraints of the six-word title and the 30-word lead. I just thought, oh this is something I really love. Here's the history there. Prior to doing a PhD in economics I was running around with poets and painters and other artists and I had the good fortune of meeting a poet called Etheridge Knight. He was a protégé initially of Gwendolyn Brooks, the poet laureate of Chicago and the United States and indeed the Black Arts Movement and Etheridge first learned poetry as a formal student from her. Anyway he became a great poet one of Robert Bly calls Etheridge Knight America's greatest oral poet ever, that’s in Bly’s opinion and Etheridge was into haiku and he knew that I had a bachelor's degree in economics and one day he said Steve, you should put haiku and economics together, that would be really cool. (Collective laughter) And I was like yeah, that would be cool and then for about 10 years, I kept thinking it would be cool and I never did it and then, funny enough you know a lot of strange things can happen in pubs and bars, right?
Bailer: You're going to share something you can for a general audience, right?
Ziliak: Exactly! And I was teaching, so this was in the early 2000s, I was teaching at the Georgia Institute of Technology in Atlanta and I had classes with up to 225 science, engineering, social science and computer science students. So I got a total of 225 kids and I was looking for ways to connect with them you know and I put on the microphone. I’m like you all, I combine humanities with sciences and statistics and I do it in my teaching and in my research and so you know I strap on the mic and I walk around that with the kids, but I still wanted a deeper connection. One day I walked into a local pub and some of my buddies were writing haiku and suddenly what Etheridge Knight had said to me back in 1991 all came rushing together and that night I wrote 22 haiku about economics.
Bailer: Wow! That was a pent-up effort!
Ziliak: In fact the very first haiku I wrote has now been published and it's shown up in a lot of different places from like The Economist magazine to Poetry mMgazine which gives me a lot of pride I have to admit.
Pennington: Yeah me too, if I were you!
Ziliak: Thank you, if you don't mind I'll tell you what that haiku is.
Bailer: Oh, please!
Pennington: Oh yeah, that’d be great!
Ziliak: It goes like this:
Mother of inflated hope,
Mistress of despair.
Pennington: Oh that's great!
Bailer: Man! I think, maybe our next contest we have to really go all haiku!
Pennington: All haiku!
Bailer: An All haiku contest!
Pennington: I'm not sure Richard would buy in on…
Bailer: Richard's not going to go in for it? What does he want, Sonnets?
Pennington: I don’t know, maybe.
Bailer: So what’s...
Ziliak: I’ll make a pitch for it because that's where the count…you know, counting and efficiency and loss functions are three things that economics, statistics and haiku share in common. So I think of that haiku form as a budget constraint, the way that economists think of prices and income and wealth as a budget constraint. But what you're trying to do is extract as much information out of that little poem, a lot like we do in statistics, trying to extract information out of let’s say, small sample data. So the other way I came into this, into Bayesian, statistics and estimation and so forth is through my criticisms of significance testing, going back to my research with Deirdre McCloskey in the 1990s. We published a paper called The Standard Error of Regression in the Journal of Economic Literature where we showed that about 70 percent of our colleagues publishing in the leading journal, general interest journal of the profession, the AER, about 70 percent of them didn't really know what significance testing could do for them. That is, Fisher's P-values, student's T-test and so on and they mistook a statistically significant result for being economically or otherwise policy important let's say and likewise if they did not find a statistical significance at a bright line level such as P<0.05 data-preserve-html-node="true" or T>1.96, then they thought they had nothing of economic importance to bother about it and ignored the parameter estimation, ignored maybe the policy opportunity or indeed maybe they ignored a great toxic social extra analogy that needs to be addressed even though not statistically significant as conventionally defined. And so the other part of that is that in microeconomics we're told that expected utility theory postulates a probability of some event relative to the utility that one would get under the realization of that event and likewise in game theory we say that there are agents who have Bayesian learning capabilities and so as the sequential game or the repeated game goes on and on, and plus whatever people can learn from it and update their beliefs the way that Bayesian probability allows them to but then when I turned around and looked at the textbooks for econometrics and statistics, it's all mill hypothesis testing and that little thing. So I say in haiku,
what are you trying to say
Bailer: Very good. I suspect that you appreciate kind of the ASA efforts with sort of the recent statement that they had made on p-values that the society has been pushing forward on kind of deeper and better understanding of these ideas. I want to give you the…
Ziliak: Absolutely. I just say that I had the great pleasure of being one of the behind the scenes contributing authors to that statement.
Pennington: Oh wow!
Ziliak: Yeah, so thank you to Ron Wasserstein and to Nicole Lazar and the Board of directors of ASA. I just I can't as Ron knows, and Barry Nussbaum and others know, I mean that is so important, that ASA statement about statistical significance and P-Values, and again I was really honored to be on that committee.
Bailer: Well that was great work by your group. So we certainly need to ask you to say what's your favorite haiku that’s statistically related, that you haven't told, that you haven't shared with us yet? What's the one, you know it's like your Encore number? So imagine that Rosemary and I are here with holding up lighters in the air!
Ziliak: OK, Here we go:
Strangling of wit
Ziliak Are your lighters still in the air?
Pennington: Mine is, I'm not sure about John! Oh my gosh! Oh man! Steve thank you so much for being here it was so much fun talking to you and I really do love your haiku. We had a really interesting mix of entries and I cannot wait for the next one, and maybe we can convince Richard to do haikus!
Ziliak: I would love it, and I’ll come back any time. I love your podcast, keep up stats and stories, and keep up your wonderful work combining humanities and statistics and the sciences, I think it's fantastic.
Pennington and Bailer: Thank you so much!
Pennington: That's all the time we have for this episode of stats and stories. Stats and stories is a partnership between Miami University's departments of Statistics and Media Journalism and Film and the American Statistical Association. You can follow us on Twitter, Apple podcast or other places you find podcasts. If you'd like to share your thoughts on the program, you can send an e-mail to firstname.lastname@example.org or check us out at our website statsandstories.net and be sure to listen for future editions of stats and stories where we discuss the statistics behind the stories and the stories behind the statistics.