More re: why “the science” is losing the public trust.
=============
It’s not a new issue! From the HomaFiles archives… circa 2015
I didn’t say it, the New Yorker magazine did, setting off a buzz in the halls of academia.
The theme of the New Yorker article –- titled “Truth Wears Off” –was that most (academic) research was flawed and not able to be replicated. This is, the results were at best true under some special circumstances at a specific point in time, but can’t be replicated. At worst, they’re just plain bull.
Hmmm.
Challenging the integrity of publication-driven academics?
Turns out that the New Yorker wasn’t the first mag on the beat.
As the post-election day hangover wears off, an examination of the mechanics behind the Clinton’s get out the vote efforts ― reaching out to Clinton voters in key states at the door, on the phone or by text messages ― reveals evidence of what appears to be a pretty shocking truth.
Clinton volunteers were inadvertently turning out Trump voters.
As the post-election day hangover wears off, an examination of the mechanics behind the Clinton’s get out the vote efforts ― reaching out to Clinton voters in key states at the door, on the phone or by text messages ― reveals evidence of what appears to be a pretty shocking truth.
Clinton volunteers were inadvertently turning out Trump voters.
The punch line: “While Donald tweets to the masses, Hillary will be precisely targeting persuadable voters.”
Advantage Hillary, right?
Maybe. Maybe not.
In an AP interview, Trump said that he “always thought that it (meaning data analytics) was overrated” and, accordingly, he’ll spend limited money on data operations to identify and track potential voters and to model various turnout scenarios that could give him the 270 Electoral College votes needed to win the presidency.
He’s moving away from the model Obama used successfully in his 2008 and 2012 wins, and the one that likely Democratic nominee Hillary Clinton is trying to replicate, including hiring many of the staff that worked for Obama in his “Victory Lab”.
=====
A data-light strategy may sound very old-school in the era of big data … especially coming from Trump …. but it reminded me of an opinion piece that Peggy Noonan wrote in the WSJ soon after Obama’s 2012 election win.
Noonan had a riff about predictive analytics that caught my eye.
It pointed out one of the downsides of predictive analytics … the craft of crunching big data bases to ID people, their behaviors and their hot buttons.
Now, researchers are trolling your web searches to auto-detect diseases.
The Washington Post recently channeled a study done by Microsoft — published in the Journal of Oncology Practice …
The essence: Microsoft’s big data analysts ID’ed folks who were querying questions like “how to treat pancreatic cancer” — hypothesizing that they might have been diagnosed with the disease.
Then, the researchers backtracked thru the prior searches done by those folks and detected a pattern of precedent queries that revolved around symptoms, e.g. abdominal swelling.
Bottom line: the researchers were able to use the inferred pattern of symptoms to early-predict a disease diagnosis for a statistically significant number people who queried symptoms.
That’s potentially big news in disease diagnosis, though doctors caution that for many diseases, the onset of patient-queried symptoms may be too late-stage for effective treatment.
====== The Microsoft query- disease analysis reminded me of how Target created some Big Data buzz for analyzing purchase patterns to ID moms-to-be.
In an AP interview, Trump said that he “always thought that it (meaning data analytics) was overrated” and, accordingly, he’ll spend limited money on data operations to identify and track potential voters and to model various turnout scenarios that could give him the 270 Electoral College votes needed to win the presidency.
He’s moving away from the model Obama used successfully in his 2008 and 2012 wins, and the one that likely Democratic nominee Hillary Clinton is trying to replicate, including hiring many of the staff that worked for Obama in his “Victory Lab”.
=====
A data-light strategy may sound very old-school in the era of big data … especially coming from Trump …. but it reminded me of an opinion piece that Peggy Noonan wrote in the WSJ soon after Obama’s 2012 election win.
Noonan had a riff about predictive analytics that caught my eye.
It pointed out one of the downsides of predictive analytics … the craft of crunching big data bases to ID people, their behaviors and their hot buttons.
Ever wonder why the gun-chewing cashier asks you for your zip code?
I naively assumed the store was just doing some kind of geo-survey … trying to figure out where their customers were coming from … how far they were driving to shop their store.
Silly boy.
CNN reports that ”Every time you mindlessly give a sales clerk your zip code at checkout, you’re giving data companies and retailers the ability to track everything from your body type to your bad habits.”
Wash Post had an interesting analysis titled “This graph shows how bad the Fed is at predicting the future
The crux of their argument: the Fed has a clear recent tendency to mis-forecast economic growth … not by a little, by a lot … forecasting almost twice as rapid growth as is ultimately realized.
For example, in 2009 the Fed was predicting 4.2 percent growth in 2011. But then in 2010 it revised that down to 3.85 percent growth. And in 2011 they revised it further to 2.8 percent growth. And when all was said and done, the economy only grew about 2.4 percent that year. The Fed projected growth almost twice as fast as what actually happened.
In the old days, folks fretted (or dreamed) about the effect of computerized automation in factories and ATMs replacing bank tellers.
According to a recent McKinsey report:
Physical labor and transactional tasks have been widely automated …
Now, advances in data analytics, low-cost computer power, machine learning, and interfaces that “understand” humans are moving the automation frontier rapidly towards “knowledge work”..
Developments in how machines process language and understand context are allowing computers to search for information and find patterns of meaning at superhuman speed.
We’re working through predictive analytics in class these days.
So, my eyes are open for articles on the subject.
Predictive analytics.
You know, the stuff that Moneyball got rolling in baseball … and Target popularized by identifying pregnant women before the women knew they were expecting.
Let’s set the stage.
The Washington Redskins have been having (another) rough season.
Veteran sportswriter Tony Kornheiser saysadvanced analytics could save the Redskins…
I didn’t say it, the New Yorker magazine did, setting off a buzz in the halls of academia.
The theme of the New Yorker article –- titled “Truth Wears Off” –was that most (academic) research was flawed and not able to be replicated. This is, the results were at best true under some special circumstances at a specific point in time, but can’t be replicated. At worst, they’re just plain bull.
Hmmm.
Challenging the integrity of publication-driven academics?
Turns out that the New Yorker wasn’t the first mag on the beat.
I didn’t say it, the New Yorker magazine did, setting off a buzz in the halls of academia.
The theme of the New Yorker article –- titled “Truth Wears Off” –was that most (academic) research was flawed and not able to be replicated. This is, the results were at best true under some special circumstances at a specific point in time, but can’t be replicated. At worst, they’re just plain bull.
Hmmm.
Challenging the integrity of publication-driven academics?
Turns out that the New Yorker wasn’t the first mag on the beat.
Wash Post had an interesting analysis titled “This graph shows how bad the Fed is at predicting the future
The crux of their argument: the Fed has a clear recent tendency to mis-forecast economic growth … not by a little, by a lot … forecasting almost twice as rapid growth as is ultimately realized.
For example, in 2009 the Fed was predicting 4.2 percent growth in 2011. But then in 2010 it revised that down to 3.85 percent growth. And in 2011 they revised it further to 2.8 percent growth. And when all was said and done, the economy only grew about 2.4 percent that year. The Fed projected growth almost twice as fast as what actually happened.
Wash Post had an interesting analysis titled “This graph shows how bad the Fed is at predicting the future
The crux of their argument: the Fed has a clear recent tendency to mis-forecast economic growth … not by a little, by a lot … forecasting almost twice as rapid growth as is ultimately realized.
For example, in 2009 the Fed was predicting 4.2 percent growth in 2011. But then in 2010 it revised that down to 3.85 percent growth. And in 2011 they revised it further to 2.8 percent growth. And when all was said and done, the economy only grew about 2.4 percent that year. The Fed projected growth almost twice as fast as what actually happened.
Speed cams are bad … AAA has done audits revealing that 1 in 10 tickets issued by them are in error … with drivers having little recourse since only the cameras are are presumed innocent until proven guilty.
Yep, they’re bad, but …
Imagine all speed limits being tightly enforced … 24 X 7.
Scary thought, right?
Here’s what will replace the speed cam … and disrupt our lives.
In the old days, folks fretted (or dreamed) about the effect of computerized automation in factories and ATMs replacing bank tellers.
According to a recent McKinsey report:
Physical labor and transactional tasks have been widely automated …
Now, advances in data analytics, low-cost computer power, machine learning, and interfaces that “understand” humans are moving the automation frontier rapidly towards “knowledge work”..
Developments in how machines process language and understand context are allowing computers to search for information and find patterns of meaning at superhuman speed.
Something caught my eye, buried deep in the weeds of the chatter re: the ObamaCare web site fiasco.
Forbes had an early-on article theorizing that a major cause of the web site problems was the Feds insistance that folks shouldn’t see potentially shocking list prices, but rather should input a lot of their private data so that they can be flashed a net price – after government subsidies.
That’s old news … and, you can believe it or not.
Here’s the passage that got me thinking:
The core problem stems from “the slate of registration systems [that] intersect with Oracle Identity Manager, a software component embedded in a government identity-checking system.”
The main Healthcare.gov web page collects information using CGI Group technology.
Then that data is transferred to a system built by Quailty Software Services.
QSS then sends data to Experian, the credit-history firm.
Hmmm
Experian – one the 3 major credit bureaus.
Why get a private sector credit bureau involved?
At first, I thought the Feds might have stumbled on a borderline brilliant idea …
Casinos have developed detailed behavioral profiles of many of their customers, based in part on information gathered though loyalty-card programs that can track slot-machine play and other-gambling activity.
The casinos use this information to tailor marketing offerings, particularly to the small minority who make up the bulk of their revenue base.
It’s called predictive analytics, and casinos have been on the leading edge.
I didn’t say it, the New Yorker magazine did, setting off a buzz in the halls of academia.
The theme of the New Yorker article –- titled “Truth Wears Off” –was that most (academic) research was flawed and not able to be replicated. This is, the results were at best true under some special circumstances at a specific point in time, but can’t be replicated. At worst, they’re just plain bull.
Hmmm.
Challenging the integrity of publication-driven academics?
Turns out that the New Yorker wasn’t the first mag on the beat.
Punch line: An increasing number of published research studies – scientific & academic – are being “retracted” because the outcomes being reported can’t be replicated or are just plain fraudulent.
The revision — .6% – may initially sound like loose change, but it’s a 25% miss.
So, economic models that operating on the original (higher) estimate have a starting point that is off by 25%.
The error compounds over time.
It’s a version of what theorists call chaos theory … how a seemingly small variation at a starting point can compound into a major effect over time.
= = = =
Side note: And, in the “new normal” economy, the downward revision was good for the stock market since it puts pressure on the Fed to continue pumping money into the economy … the bulk of which is flowing straight to the stock market.
Wash Post had an interesting analysis this week titled “This graph shows how bad the Fed is at predicting the future
The crux of their argument: the Fed has a clear recent tendency to mis-forecast economic growth … not by a little, by a lot … forecasting almost twice as rapid growth as is ultimately realized.
For example, in 2009 the Fed was predicting 4.2 percent growth in 2011. But then in 2010 it revised that down to 3.85 percent growth. And in 2011 they revised it further to 2.8 percent growth. And when all was said and done, the economy only grew about 2.4 percent that year. The Fed projected growth almost twice as fast as what actually happened.
Everybody knows that Blackjack is a game of probabilities and that card-counting can get you kicked out of casinos – because it helps slightly with the odds.
Did you know that math and statistics can also improve your odds in Monopoly?
Business Insider posted a fun (and thorough) pitch re: how to win in Monopoly … a great practical (?) application of math and statistics.
Here are a couple of takeaways and a link to the entire pitch … worth browsing, even if you’re not a Monopoly aficionado.
In a prior post Feds: “Hire ex-cons … we do” … say, what? … we reported that Feds are hiring ex-cons into the State Department … and pressuring private companies to hire them, too.
Sounds risky to me, but what if there was some objective way to cut the risk … to determine the likelihood that an ex-con would (or would not) go straight.
There is a way.
Some courts and parole departments are using predictive analytics to help decide who belongs in prison.
Yesterday we posted about a company developing algorithms that comb through key factors including content of posts, and location, among others
….. to provide a very to develop a identify and “unify social profiles” for users who may be using different names or handles on each of their social networks.
The post elicited strong interest and 2 replies that I want to highlight.
The first is from Niv Singer, Chief Technology Officer at Tracx … the man, and the company referenced.
In the old days, folks fretted (or dreamed) about the effect of computerized automation in factories and ATMs replacing bank tellers.
According to a recent McKinsey report:
Physical labor and transactional tasks have been widely automated …
Now, advances in data analytics, low-cost computer power, machine learning, and interfaces that “understand” humans are moving the automation frontier rapidly towards “knowledge work”..
Developments in how machines process language and understand context are allowing computers to search for information and find patterns of meaning at superhuman speed.
Short answer: you have new cookies installed on your computer or have old cookies modified … whether you know it or not … and you then spew crumbs all over the Internet … letting companies track you, profile you, and hard sell you stuff.
Here’s a visual of what a couple of clicks can do … each dot represents a site or company that can grab your information … just because you innocently clicked.
Later we’ll explain the graphic and what’s going on.
Ever wonder why the gun-chewing cashier asks you for your zip code?
I naively assumed the store was just doing some kind of geo-survey … trying to figure out where their customers were coming from … how far they were driving to shop their store.
Silly boy.
CNN reports that ”Every time you mindlessly give a sales clerk your zip code at checkout, you’re giving data companies and retailers the ability to track everything from your body type to your bad habits.”
Peggy Noonan has a piece in the WSJ today that I almost skipped.
You know, another “Is Obama a Lame Duck?” piece.
Buried in the column was a riff about predictive analytics that caught my eye.
It pointed out one of the downsides of predictive analytics … the craft of crunching big data bases to ID people, their behaviors and their hot buttons.
These days the business and management science worlds are focused on how large datasets can decode consumers’ behavior patterns … enabling marketers to laser-target high potential prospects with finely-honed messages, offers, and “attention”.
“Big data” … becomes problematic when it adheres to “data fundamentalism” … the notion that correlation always indicates causation, and that massive data sets and predictive analytics always reflect objective truth … that “with enough data, the numbers speak for themselves.”
Big data has hidden biases in both collection methods and analysis …
Punch line: Online retailers are using sophisticated analytics and web tracking methods to tailor their offerings… and to get folks to pay higher prices.
To get the lowest prices: (1) Use a PC (not Mac or iPad), (20 don’t sign on from a ritzy location, (3) pass thru a price-shopping site on your way to the destination site, (4) ask to see items sorted by price — from low to high, (5) check out at least one cheap item — maybe even put in your cart — then delete it later
Retailers are becoming bigger users of so-called predictive analytics, crunching reams of data to guess the future shopping habits of customers.
The goal is to tailor offerings to people believed to have the highest “lifetime value” to the retailer.
Online, seemingly innocuous information is available to predict shoppers’ tastes and spending habits.
For example, The average household income for adult owners of Mac computers is $98,560, compared with $74,452 for a PC owner.
Drilling down, Orbitz has found that people who use Apple spend as much as 30% more a night on hotels, so the online travel agency is starting to show them different, and sometimes costlier, travel options than Windows visitors see.
More specifically …
Mac users on average spend $20 to $30 more a night on hotels than their PC counterparts, a significant margin given the site’s average nightly hotel booking is around $100
Mac users are 40% more likely to book a four- or five-star hotel than PC users,
When Mac and PC users book the same hotel, Mac users tend to stay in more expensive rooms.
Other factors that have influence over results include
A user’s location (e.g. geo-targeting high wealth zip codes)
A shoppers history on the site (e.g. purchases at list price or at discounts).
The referring site (e.g. Kayak delivers price-sensitive shoppers to travel sites)takes those properties into account.
The Homa Files 