Posts from November 2018 (10)

Team Ratings for Round 9

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 48 matches played, 38 of which were correctly predicted, a success rate of 79.2%.
Here are the predictions for last week’s games.

Predictions for Round 9

Here are the predictions for Round 9. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Team Ratings for Round 10

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 63 matches played, 52 of which were correctly predicted, a success rate of 82.5%.
Here are the predictions for last week’s games.

Predictions for Round 10

Here are the predictions for Round 10. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Two press releases from Stats NZ today

2018 Census data release delayed

Why the 2018 Census data release is delayed

Brian Wansink, a prominent food researcher from Cornell, was forced to retire earlier this year. Andrew Gelman has some good perspectives. Wansink’s research was on contextual effects on eating — eg, the impact of plate size — and a bunch of these papers have now been retracted.

This week, Wired has a Thanksgiving-themed article about his research. It includes this quote from an email he sent to colleagues ahead of his retirement

“We may believe that our papers have been unfairly retracted. But what they can’t retract is the impact these have had on people’s lives and the impact they will continue to have.”

That’s a perfect summary of the problems with studies that over-promise and are over-publicised.  Whether the research was done well or not, it’s going to stick.  Subsequent developments — whether modifications, replication failures, or retractions — never get the impact of the original claim.

• “Rather than assume algorithms will produce better outcomes and hope they don’t accelerate discrimination, we should assume they will be discriminatory and inequitable unless designed specifically to redress these issues.” Lucy Bernholz
• ” Introduction of [a predictive risk screening tool] resulted in a statistically significant increase in emergency hospital admissions and use of other [National Health] services without evidence of benefits to patients or the [National Health Service].” In the academic journal BMJ, so a bit more technical
• Why the NY Times map of the US election results is so good: a Twitter thread
• Stacey Kirk in the Sunday Star-Times on the campaign to get Pharmac to pay for one of the most expensive drugs in the world.
• Interesting interactive in the Herald about quality-of-life and work in NZ cities.  It’s very economist in style.  That’s true on the good sense that it appreciates high house prices are a signal that lots of people want to live somewhere and low house prices are a signal that lots of people don’t.  It’s also true in the bad sense that there some places where not many people want to live, but the people who do live there really like it — and this sort of analysis suppresses that variation in preferences.
• Interesting book on data science and data use: “Data Feminism”

The Herald has a story from the Washington Post on an “AI” screening tool for babysitters, that allegedly uses both computer vision and text processing to screen social media for risk factors.   Here are three quotes from it:

1. A company co-founder says

Parents, he said, should see the ratings as a companion that “may or may not reflect the sitter’s actual attributes.”

But the danger of hiring a problematic or violent babysitter, he added, makes the AI a necessary tool for any parent hoping to keep his or her child safe.

The first thing to note about this is that you could make the same claims about astrology or handwriting analysis or a tarot reading. There’s no quantitative information about accuracy given, and it’s hard to see how the company could even know much how accurate its ratings were, or how biased. It’s not even for sure that the risk rating is positively correlated with risk to kids; the company seems careful not to make even a claim this weak.

2.

Parents could, presumably, look at their sitters’ public social media accounts themselves. But the computer-generated reports promise an in-depth inspection of years of online activity, boiled down to a single digit: an intoxicatingly simple solution to an impractical task.

If the algorithms actually predicted risk better and were less biased than typical employers there might be an advantage to this: your social media would be shared with a faceless US company rather than your potential employer, so there might be less actual privacy invasion — after all, some faceless US companies already have your social media. It might be less embarrassing than your boss knowing what your favourite member of the appropriate sex calls you. The computer could also be set up to ignore irrelevant information like whether you talk about your sexual orientation online.   With the setup as it is, that’s not the case, and one of the biggest risks is the completely unfounded appearance of both accuracy and objectivity — “mathwashing” as the jargon puts it.

3.

Where she lives, “100 per cent of the parents are going to want to use this,” she added. “We all want the perfect babysitter.”

One of the significant risks of automating human judgement is that it can go viral. There’s a limit to how well ordinary human prejudices can scale — you’ve got some chance of finding someone who has different biases. The prejudices of one computer checklist, though, can keep someone completely out of an employment sector.

Team Ratings for Round 8

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 42 matches played, 32 of which were correctly predicted, a success rate of 76.2%.
Here are the predictions for last week’s games.

Predictions for Round 8

Here are the predictions for Round 8. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Team Ratings for Round 9

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 56 matches played, 46 of which were correctly predicted, a success rate of 82.1%.
Here are the predictions for last week’s games.

Predictions for Round 9

Here are the predictions for Round 9. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Yesterday and today we had talks by our postgraduate students about their short research projects

• An investigation into criticism of the FST software. (forensic DNA analysis: you want it to be right)
• Comparison of accuracy between different algorithms for solving the Normal Equations in Regression Analysis (It’s hard to get big rounding errors nowadays)
• Multi-catchment Streamflow Modelling by Reduced-rank Regression (improving hydropower modelling)
• A Study on Equity in Academic Outcomes in First Year Statistics Courses (Could do better)
• Costs and Financing of Routine Immunisation (estimating cost components in low-middle income countries)
• Statistical Examination of the Relationship between Maternal Diet, Metabolome and Gestational Diabetes Mellitus (predicting diabetes is hard, especially in the future)
• Robustness of Spatial Capture-Recapture Models to Misspecified Detection Functions (if you’re listening for whales or gibbons, how close do you need to be?)
• An Examination of Participant Perspectives on the Scampy Tool (teaching the ideas of randomness at an introductory level)
• Who lives in deprived places? The association between individual and area level socioeconomic position (guessing someone’s income from where they live isn’t all that reliable)
• Is LIBS a reliable technology for the forensic analysis of glass? (more forensics, but with lasers)
• De-batching data from a complex experiment (life would be simpler if you didn’t have to worry about lab ‘batch effects’ when studying ocean acidification)
• Optimal path in random graphs (maths about really big networks)
• Spatial Distribution of fish on the Chatham Rise (if you want to count them, you need to know where to look)
• Exploring climate variables (in particular, extreme values like hottest place, rainiest day)
• Interactive Tools for Climate Data (for browsing through NZ historical weather data)
• How old is that mud: Convex Biclustering Applied in Tephrostratigraphy of the Orakei Basin (Looking for volcanic ash layers in mud samples)
• Comparison of Methods for Inferring Granger Causality (time series techniques for economists)
• Predicting Patronage (how does Patreon support vary over time, and can you predict it?)
• Expected Information Matrices for Some Poisson Variants. (Calculations and software for some new counting models)

From Axios

The wealth of the world’s billionaires rose by $1.4 trillion in 2017, the largest annual increase ever.

The details: Nearly all of that increase was driven by the Asia-Pacific region, and specifically China, where billionaire wealth rose 39%.

The graph is not well designed for illustrating the claim about billionaires: in a stacked bar chart like this it is easy to compare the first level, the sum of the first two, and the sum of all three, but not the top level.  If the chart had been stacked the other way up it would be more obvious that it doesn’t seem to go along with the claim.

Measuring the height of the bars in MacOS Preview (because it’s not possible to read the graph to high enough precision) I get 35 pixels increase in the Asia-Pacific region and 46 in the rest of the world, so the growth in the Asia-Pacific region actually is less than half the total growth.  Estimating the total growth by counting pixels does agree with the $1.4 trillion total, so what has gone wrong?

Clicking through to the UBS/PwC press release gives a bit more detail:

Chinese billionaires increased in number to 373 in 2017 from 318 in 2016 and their wealth rose by 39 percent to USD 1.12 trillion

So, the net increase in total wealth for Chinese billionaires is about $0.44 trillion. That hardly qualifies as “nearly all” of the increase for the Asia-Pacific region, let alone for the world

The original source is the the UBS/PwC Billionaires report. Their graph is better (apart maybe from the second y-axis). Not only is it the right way up for looking at increases in Asia vs the rest of the world, but it’s got tick marks on the right-hand axis where they’re actually useful. And it shows some historical context.

Normally I wouldn’t go to this much effort for a business news item — but the Axios Edge newsletter where I read this is edited by Felix Salmon, who is usually better at the difference between “nearly all” and “nearly half”

P.S. Yes, I did think of titling this post Crazy Rich Asians. No, I’m not sorry I didn’t.