Here is a quote I liked from the blog post “Statistics: Losing Ground to CS, Losing Image Among Students”:

(Recently I was pleased to discover–“learn,” if you must–that the famous book by Hastie, Tibshirani and Friedman complains about what they call “hype” over neural networks; sadly, theirs is a rare voice on this matter.)

Cheers.

 

This is the first of the blog posts in my “backblog” pertaining to #jsm2014.  

My dissertation work was in statistical disclosure control and the post-doc work was in genetics.  Almost immediately after starting my post-doc, I realized that privacy issues and genetic data seem to go hand in hand.  (I recently submitted and R21 to the NIH about using synthetic data to protect privacy in genome-wide association data.  It was not funded.  I will re-submit.) Anyway, the point is when I saw this session entitled “Genomic Privacy: Risk and Protection Methods”, I absolutely had to go to it.  The talks were all fantastic.

The first speaker was Guy Rothblum from Microsoft who was presenting work with Cynthia Dwork on an extension to differential privacy called concentrated differential privacy (CDP).

Here is what they say about CDP in their abstract:

We introduce Concentrated Differential Privacy (CDP), a relaxation of Differential Privacy geared towards improved accuracy and utility.

Like Differential Privacy, Concentrated Differential Privacy is robust to linkage attacks and to adaptive adversarial composition. In fact, it composes as well as Differential Privacy, while permitting (significantly) smaller noise in many setting.

This seems like a good step forward for differential privacy and attempts to address some of the very real issues with the method.

Guy’s introductory slides were a fantastic explanation of the problem at hand and he made a lot of really interesting points. One of the reports he mentioned in his intro slides was the big data review by the federal government.  I haven’t finished reading all of this yet, but what I have read is really interesting.  Check out the report here: Big Data Review (This report deserves its own blog post.)

Here is a list of some other points that I write down as quickly as I could during his talk:

• Anonymization (removing identifying attributes) does not seem robust in the context of big data. Defeated by the presence of big data.  (Netflix Prize as an example of failed anonymization.)
• Concerns 1.) linkage attacks via partial information (ALL adversaries have partial information about us)
• Concern 2.) Composition: Each query is private in isolation, but not necessarily in multiple analyses
• Concern 3.) “Just release statistics”: Attacks include differencing, big band attacks. (e.g. Query 1: How many sickle cell individuals in DB Query 2: How many sickle cell individuals not names Guy Rothblum. Difference of these queries yields the status of Guy Rohtblum.)
• Intuition of differential privacy: “Bad things can happen happen but not because you participate in the data base.”
• Advantages of differential privacy: 1.) quanitifiable 2.) handles linkage attacks/auxiliary data
• Concentrated differential privacy improves the noise addition by an order of magnitude.  Better accuracy, mildly relaxed privacy.
• “A social choice must be made” This is a great point.  Once we can quantify privacy, which we don’t all agree on yet, we need to have a discussion about how much privacy we want.

The second speaker was Bradley Malin from Vanderbilt who spoke about “Anonymization of Phenotypic data to support genotype association studies”

Some bullet points from his talk:

• Two quotes I took away from his first few slides were “Hurdle not Fort Knox” and “Possible doesn’t imply probable”.  In terms of the first quote my boss always describes this in terms of breaking into a house: Just because it’s illegal to break into someone else’s home, doesn’t mean we don’t lock our doors.  But you don’t necessarily need bars on the windows either.  We can’t simply rely on the law to deter adversarial data users, but we also don’t need to go over board.
• “Often we use very strong adversary models. But almost perfect results can be achieved…. in the real world.  We must be ‘reasonable and practical‘” I am totally guilty of this.  A few of my articles on the topic make very strong assumptions about what the adversary knows (often “worst case scenario).  My more recent papers are relaxing these worst case scenario assumptions.
• Examples of things that could potentially identify an individual: demographics,diagnosis codes, laboratory, DNA, location visits, movie reviews.
• Malin introduces a procedure named UGACLIP for anonymization in a GWAS setting.
• Malin talked about how stakeholders should some how participate in the decision as to how much privacy they think they should have, but many people have no idea how to interpret the numbers.  (i.e. Average person on the street doesn’t have any idea how secure 5-anonymization is)
• When sharing data with NIH the generally accepted value of k (as in k-anonymzation) is 5.

 

Speaker 3 was Fei Yu of Carnegie Mellon who presented joint work with Stephen Fienberg. Fei spoke about “Scalable Privacy preserivng data sharing methodologies for GWAS.” (Full article on arXiv).

• One of the big privacy concerns of GWAS is that even aggregate statistics from a GWAS (MAF, χ² statistics, regression coefficients) do not provide perfect privacy.  Homer et al. (2008) showed that an intruder may be able to infer that someone has participated in a GWAS.  This caused the NIH to review its data sharing policy of GWAS data.
• Fei presented work on how to share the top M SNPs (in terms of their significance) and achieve ε-Differential Privacy.

Finally, Hae Kyung Im of the University of Chicago spoke about the “On Sharing Quantitative Trait GWAS Results in an Era of Multiple-Omics Data and the Limits of Genomic Privacy“.  Since I am now a resident of the midwest, I meant to try to meet Hae Kyung Im, but at the end of the session I bumped into someone else and ended up missing her (this happens all the time at JSM.  You try to do one thing, but then you bump into someone you haven’t seen in 5 years.)

• “For full advantage, broad sharing of data of results is needed; Must be careful about privacy.”  I totally agree with this.  There is so much potential benefit to sharing this type of data that we can’t just lock it up in a database and throw away the key.
• “Summary studies are considered safe, BUT with GWAS studies we may have millions of SNPS”.  With big data our previous ideas about what is safe to release need to be re-avaluated.  She again cited Homer et al. (2008) noting based on that article: Even if the DNA sample was a mixture of 1000 or more individuals, they were able to determine with high accuracy whether they were in the sample or not.  This is “Great for forensics, but has consequences fro GWAS”.
• The question she posed in terms of GWAS is basically “Can we publish regression coefficients?” I was going to try to summarize her results, but I can’t interpret what I wrote about her results, so I’ll just wait until I get her slides. (I tweeted (Twitter is awesome) at her, hopefully she will be kind enough to share them.)

A discussion then ensued, which led Stephen Fienberg to make (roughly) the following statement:

I think that there is a misconception.  In very high dimensional problems as in GWAS with the auxiliary data every individual is unique.  You must find something to share.  We cannot go on saying we can’t share anything.  If I were on your IRB there is no issue with the faculty doing this.  The only issue is what do they publish on what they have done.  And I don’t think there is an IRB that understands how to do that in the entire country.

One final thought: I was struck by how few people were in this session.  This will be is already a big deal.  Right now with just a single hair, someone can reproduce your face up to a family resemblance.  Who knows what we’ll be able to do in 5, 10 or 50 years?  Privacy is a big deal.

Cheers.

The 2014 NFL season is almost here.  That means it’s time for my NFL preview  (2013 NFL Season Preview).  I’ve updated and tweaked my model this year so it’s hopefully better than last year. Below you can read my predictions for the 2014 season. Or if you just want to see the numbers and skip my terrible comparisons of NFL divisions to different countries economies, jump to the bottom. Cheers.

AFC

East

The AFC east the story of America and the Patriots are the 1%-ers.  The rest of the division is the 99%.  In that, I’m 99% sure that none of these other teams will make the play-offs this year.  Pick: Patriots win the division by 7 games.

North

If the AFC east is all about inequality, the the AFC North is a socialist paradise.  Like the Finland.  I have 3 teams going 9-7 and all making the play-offs with Cleveland being the exception.  (Cleveland is like a poor guy in Finland; You won’t read that sentence anywhere else.  I promise you that.)  Though if this division ended with four 8-8 teams I wouldn’t be surprised at all.  Pick: Baltimore (Or Pittsburgh) (Or Cincinnati) (In Finland, everyone makes the playoffs).  

South

Want another economy comparison?  Well you’re getting one.  The AFC South is like India.  The have people who are rich (Houston), but not rich, rich like America (New England), they have a developing middle class that is good not great (Indianapolis and Tennessee) and absolute, abject poverty at the bottom (Jacksonville).  Boom.  That actually sort of makes sense, right?  Pick: Houston.  Because we all know they weren’t really 2-14 bad.  

West

I’ve got nothing left.  Wait.  I do.  The AFC west is like China.  You have the super rich (Denver), a developing middle class (Kansas City and San Diego) and a terrible lower class (Oakland) that is poor, but not as poor as India’s poorest (Jacksonville).  Pick: Denver.  Because (if I was from ESPN) PEYTON! But really because they are the best TEAM in the AFC West.  

NFC

East

Alright.  I guess I’m committed to this now.  The NFC East is….Denmark.  One of the most socialist countries in Europe with one of the lowest poverty rates of all OECD nations according to Wikipedia.  Philadelphia finishes 10-6, everyone else finishes 8-8.  And I’m sure Dallas will have a chance to win and get in in their last game of the season and Romo will throw an interception to lose for his team because everyone loves that narrative. (I’m not sure how exactly that is like Denmark, but it is.) Pick: Philadelphia.  Because someone has to win.  

North

Let’s go ahead and say the NFC North is like Brazil.  4 fan bases that couldn’t possibly be sadder when their team let’s them down.  Just like Brazil and the World Cup.  (This is officially off the rails)   Pick: Green Bay.  Because they are publicly owned.  

South

The NFC South is like…France?  Because there are a lot of teams that want to make the play-offs but there just aren’t enough spots (NFC South Teams:French Youth::NFL Play-off spots:French Jobs). And New Orleans.  Pick: New Orleans.   Because they are the most French.

West

The NFC West is clearly Russia.  I can’t do this anymore.  Pick: San Francisco.  Slightly easier schedule than Seattle.  

So who’s gonna make the play-offs?

Pretty boring in the AFC.  I’m picking the exact same 6 teams to make the play-offs as I did last year (New England, Denver, Houston, Baltimore, Pittsburgh, Cincinnati).  (What’s the definition of insanity again? Doing the same thing….)  In fact, the only thing I switched in the AFC was the seeds of the 5 and 6 teams (I swapped Pittsburgh and Cincinnati). In the NFC, I’m obviously picking San Francisco and Seattle to make the play-offs (as is just about everyone else) and one of these teams will be the 1 seed, with the edge to San Francisco who plays the (slightly) easier schedule.  Along with those two, I’m taking New Orleans, Green Bay, and Philadelphia to win their divisions.  And finally, as the 6 seed in the NFC I’m taking….(drumroll please)….Detroit!  I have them squeaking in at 9-7. Finally, I’m taking San Francisco over New Orleans in the NFC title game and New England over Denver in the AFC.  This leads to a Patriots vs 49ers Super Bowl with the Patriots prevailing 24-23.  (I know.  Boring right?  Same pick as last year.)

Ranks

Retro is based only on games played in 2013 and heavily weighs strength of schedule.  New Orleans is ranked number 1 based on this as they finished 11-5 and played 10 teams who finished 8-8 or better.

Prosp is based on 4 years of data weighted fro recency.  It’s based on expected points with New England ranked number 1 in this measure.

Both Retro and Prosp are framed in terms of the probability of defeating and average team.

predMargin column is the average margin of victory if each team played every other team at home and away.  Denver is ranked number 1 in this measure due to their ability and willingness to score a ton of points.

Team Retro Prosp PredMargin New England 70 79 4.52 Seattle 68 70 3.66 San Francisco 58 69 3.72 Denver 64 68 4.56 New Orleans 71 67 3.17 Philadelphia 44 62 1.45 Carolina 61 61 1.58 Green Bay 46 61 2.93 Chicago 50 56 1.11 Houston 29 56 0.89 Washington 33 55 -0.49 Atlanta 40 54 0.56 Detroit 48 53 1.20 Kansas City 50 53 -0.47 San Diego 52 52 0.27 Cincinnati 66 51 0.95 Buffalo 47 47 -1.93 Baltimore 56 46 0.53 Indianapolis 58 45 -1.23 Pittsburgh 51 43 0.23 Dallas 54 43 0.19 St. Louis 51 43 -2.43 Tampa Bay 40 42 -1.86 Minnesota 49 41 -1.78 NY Giants 45 41 -0.32 Oakland 37 39 -3.47 NY Jets 54 38 -2.25 Arizona 58 38 -1.74 Tennessee 41 36 -2.32 Miami 44 32 -1.79 Cleveland 39 31 -2.92 Jacksonville 26 24 -6.52

Playoff Probabilities

Team WinDivison MakePlayoffs MakeSuperBowl WinSuperBowl Arizona 0.18 1.10 0.06 0.00 Atlanta 10.40 23.60 2.28 0.98 Baltimore 36.54 63.96 6.16 2.10 Buffalo 0.14 6.00 0.28 0.10 Carolina 18.88 37.98 4.52 1.98 Chicago 14.42 29.64 3.14 1.48 Cincinnati 29.70 58.72 5.78 2.26 Cleveland 1.12 4.32 0.30 0.00 Dallas 20.62 28.48 3.16 0.98 Denver 95.04 99.38 28.80 15.20 Detroit 20.32 38.96 5.08 2.06 Green Bay 64.70 78.06 15.68 6.92 Houston 86.98 92.04 14.92 6.32 Indianapolis 8.34 21.46 1.38 0.48 Jacksonville 0.00 0.00 0.00 0.00 Kansas City 1.10 22.58 1.16 0.48 Miami 0.32 10.10 0.70 0.16 Minnesota 0.56 1.76 0.08 0.08 New England 99.42 99.78 29.82 19.54 New Orleans 70.26 82.94 17.06 9.36 NY Giants 15.38 21.46 2.20 0.70 NY Jets 0.12 3.32 0.26 0.08 Oakland 0.00 0.68 0.02 0.00 Philadelphia 48.68 57.48 8.76 4.54 Pittsburgh 32.64 61.18 6.26 1.84 San Diego 3.86 43.02 3.20 1.26 San Francisco 51.00 88.36 18.70 10.36 Seattle 48.76 86.70 16.76 9.36 St. Louis 0.06 0.50 0.02 0.00 Tampa Bay 0.46 1.16 0.06 0.02 Tennessee 4.68 13.46 0.96 0.24 Washington 15.32 21.82 2.44 1.12

Projected Records

Team – (Median wins) expected wins

AFC East

New England – (13-3) 13.044

Miami – (6-10) 6.344

Buffalo – (6-10) 5.905

NY Jets (5-11) 5.329

AFC North

Baltimore (9-7) 9.192

Pittsburgh – (9-7) 9.129

Cincinnati – (9-7) 9.041

Cleveland – (5-11) 5.328

AFC South

Houston (11-5) 10.679

Indianapolis – (7-9) 7.114

Tennessee (7-9) 6.623

Jacksonville (2-14) 2.234

AFC West

Denver – (13-3) 12.636

San Diego – (8-8) 8.334

Kansas City – (7-9) 7.369

Oakland – (4-12) 4.213

NFC East

Philadelphia (10-6) 9.579

Dallas (8-8) 8.256

NY Giants (8-8) 7.801

Washington (8-8) 7.75

NFC North

Green Bay (11-5) 10.659

Detroit (9-7) 9.095

Chicago (9-7) 8.505

Minnesota (5-11) 5.352

NFC South

New Orleans (11-5) 10.990

Carolina (9-7) 8.909

Atlanta (8-8) 8.227

Tampa Bay (5-11) 5.227

NFC West

San Francisco (12-4) 11.594

Seattle (11-5) 11.449

Arizona (5-11) 5.312

St. Louis (5-11) 4.781

Projected Playoffs

AFC

1. New England

2. Denver

3. Houston

4. Baltimore

5. Pittsburgh

6. Cincinnati

NFC

1. San Francisco

2. New Orleans

3. Green Bay

4. Philadelphia

5. Seattle

6. Detroit

Projected Wild Card Round

AFC

Houston beats Cincinnati  22-20

Baltimore beats Pittsburgh 22-20

NFC

Green Bay beats Detroit 27-23

Seattle beats Philadelphia 23-22

Projected Divisional Round

AFC

New England beats Baltimore 28-22

Denver beats Houston 27-21

NFC

San Francisco beats Seattle 21-19

New Orleans beats Green Bay 27-24

Projected Conference Round

AFC

New England beats Denver 29-27

NFC

San Francisco beats New Orleans 24-22

Super Bowl

New England beats San Francisco 24-23

Season Long Bets

Win Totals

Arizona Under 7.5 +120

Cleveland Under 6.5 +115

Denver Over 11.5 +105

Houston Over 7.5 +125

Indianapolis Under 9.5 +110

Jacksonville Under 4.5 +145

NY Jets Under 7 +105

Kansas City Under 8.5 +145

Miami Under 7.5 -130

New England Over 10.5 -210

St. Louis Under 7.5 +135

Tampa Bay Under 7 +125

Win Division

Philadelphia +130

Green Bay -130

New England -320

Denver -300

Make Playoffs

AFC

Cincinnati +150

Baltimore +220

San Diego +300

Houston +280

NFC

Seattle -300

San Francisco -250

New Orleans -140

Carolina +235

Detroit +270

Miss Playoffs

Indianapolis -140

Crazy Long Shot Super Bowl Match-Up

Carolina vs Houston +70000