Stats in the Wild

recent posts

about

  • NCAA Basketball Sweet Sixteen edition (in the wild)

    Well there are 16 teams left and my bracket is in shambles.

    Let’s review the predictions from last week.
    My failures:
    Two of my final four teams are gone (Villanova and Kansas) including my champion (Kansas) so I won’t win any office pools this year.

    My triumphs:
    -A lot of my predictions from my tournament preview came through including both of my predicted lower seed first round locks (Northern Iowa and Missouri).

    -I had Northern Iowa in my top 25 . Certainly not higher than Kansas, but definitely a top 25 team.

    -My model ranked Cornell 18th and I ignored it figuring it was a fluky part of the model. (Similar to how I have Oral Roberts ranked 8th and Sam Houston State ranked 2nd using the raw data.) I guess Cornell really is that good. This win over Wisconsin should shoot them up in my rankings when they come out tomorrow.

    Some observations:
    -Washington is really bad. They beat a mediocre Marquette team and then a New Mexico team that lost to San Diego State in its conference tournament. West Virginia is going to massacre Washington. (Note: Marquette lost 51-50 to DePaul this season. DePaul was 1-17 in the Big East. Yikes.)

    -The Kentucky-Cornell game is going to be really interesting. Kentucky is loaded with super talented Freshman who are thinking about the NBA and millions of dollars and Cornell is stacked with seniors who are thinking about grad school next fall.

    My picks:
    -Kentucky over Cornell: Cornell keeps it close in the first half, but Kentucky’s superior talent takes over. They win by 10.
    -West Virginia over Washington: West Virginia is going to kill them. I see them winning by 15-20 points.
    -Duke over Purdue: This is an interesting match-up. Purdue had thoughts of a number 1 seed going into their conference tournament and they ended up with a 4. I’ll be interested to see how they play in this one.
    -Baylor over Saint Mary’s: Baylor has beaten a 14 and an 11 seed. Saint Mary’s is a 10. I’m not sure what that means, but the total of the seed of Baylor’s first three opponents is 35. That has to be the highest total of a first three opponents, right? Anyway, this one will be close. Baylor by 3.
    -Northern Iowa over Michigan State: I had Northern Iowa in top 25 at the beginning of the tournament and they are still there. I had Michigan State out of my top 25. Northern Iowa by 7. (Michigan State has made the Sweet Sixteen three years in a row.)
    -Tennessee over Ohio State: How many games does Ohio State have to win in a row to convince me they are for real? 1 more. But I doubt it. Tennessee by 10.
    -Syracuse over Butler: Syracuse by 5…..unless Shelvin Mack hits a million threes again. Then who knows.
    -Kansas State over Xavier: Kansas State looks good. really good. (Note: Xavier has made the Sweet Sixteen three years in a row.)

    Final Four picks:
    Kentucky, Duke, Syracuse, and……(wait for it)…………Northern Iowa. Northern Iowa beat Kansas, surely they can beat Michigan State, and Ohio State or Tennessee, right?

    Finals:
    Syracuse vs Kentucky

    Champion:
    Kentucky 72-68

    Can’t wait until next Monday when I get to see just how wrong I am again.

    Cheers.

  • NCAA Basketball top 25 (in the wild)

    (Bold Indicates Sweet Sixteen team.)
    StatsInTheWild top 25:

    1. Kansas
    2. Kentucky
    3. Syracuse
    4. West Virginia
    5. Villanova
    6. New Mexico
    7. Duke
    8. Temple
    9. Kansas State
    10. Georgetown
    11. Baylor
    12. Pittsburgh
    13. Purdue
    14. Texas A&M
    15. Tennessee
    16. Texas
    17. Butler
    18. Vanderbilt
    19. Ohio State
    20. Marquette
    21. Maryland
    22. Richmond
    23. Northern Iowa
    24. Xavier
    25. BYU

    Inexplicable regular season losses:
    Penn beat Cornell 79-64
    Indiana beat Pittsburgh 74-64
    Brown beat Princeton 57-54
    Evansville beat Northern Iowa 55-54

    Cheers.

  • NCAA Basketball (in the wild)

    It’s that magical time of year again. The three weeks the rest of the country and I care about college basketball. Check out the StatsInTheWild NCAA basketball top 25.

    So here it is. The StatsInTheWild annual NCAA tournament preview.

    Teams that should have gotten in but didn’t:
    Seton hall – I realize it’s hard to take a team that went 9-9 in their conference, but it’s the big east. it really is that good.
    Virginia Tech – They should have been in easy. Although 2 losses to lowly Miami, including once in the ACC tournament is really bad.
    Mississippi State – A good regular season and a very good run in the SEC tournament. Should have been in.

    Teams that should not have gotten in:
    Minnesota – They lost to Michigan twice in February and they lost to Indiana who was 4-12 in the conference. And they were only 9-9 in the conference. The big 10 is the most over rated conference in football and basketball.
    UNLV – Third in the Moutain West gets in but third in the ACC doesn’t? This was a bad at large bid.
    Wake Forest – They finished 6th in the ACC at 9-7 in conference. How did Virgina Tech not get in again?
    Georgia Tech – They finished 7th in the ACC at 7-9 in conference. How did Virgina Tech not get in again?

    Best 16 seed: Lehigh
    Best 15 seed: UC-Santa Barbara
    Best 14 seed: Sam Houston State

    Worst 1 seed: Duke
    Worst 2 seed: Ohio State
    Worst 3 seed: Pittsburgh
    Worst 4 seed: Wisconsin

    Most likely first round upsets:
    (14) Sam Houston State over (3) Baylor
    (12) New Mexico State over (5) Michigan State
    (13) Wofford over (4) Wisconsin
    (11 )Old Dominion over (6) Notre Dame

    Most likely long shot upset:
    (15) UC-Santa Barbara over (2) Ohio State

    Lower seed lock:
    (10)Missouri over (7) Clemson
    (9) Northern Iowa over (8) UNLV

    Sweet Sixteen:
    All the one’s, two’s, and three’s along with (4) Maryland, (4) Butler, (5) Texas A and M, and (5) Temple.

    Elite 8:
    All the number one seeds and all of the number 2 seeds except Ohio State. Georgetown gets in.
    So that’s (1) Kentucky, (1) Duke, (1) Syracuse, (1) Kansas, (2) Kansas St, (2) Villanova, (2) West Virginia, and (3) Georgetown.

    Final 4:
    (1) Kansas, (1) Kentucky, (1) Syracuse, (2) Villanova

    Finals:
    (1) Kansas vs (1) Kentucky

    Champion:
    (1) Kansas over (1) Kentucky 68-66

  • LaTeX and WordPress (in the wild)

    Apparently, you can use LaTeX in wordpress. Alright, he is a practice problem I was working on for my exam in two weeks.

    Let \Sigma=(1-\rho)I_{k}+\rho J_{k} with 0 \le \rho \le 1 where I_{k} is a k x k identity matrix and J is a k x k matrix of ones. Find the eigenvalues of \Sigma.

    We seek \lambda such that det(\Sigma - \lambda I)=0 where det(A) is the determinant of a matrix A

    det((1-\rho)I_{k}+\rho J_{k}-\lambda I_{k})=0
    det((1-\lambda-\rho)I_{k}+\rho J_{k})=0
    det(\rho(J_{k}+\frac{1-\rho-\lambda}{\rho}I_{k}))=0
    det(J_{k}+\frac{1-\rho-\lambda}{\rho}I_{k}))=0
    det(J_{k}-(\frac{-(1-\rho-\lambda)}{\rho}I_{k}))=0
    det(k(\frac{J_{k}}{k}-(\frac{-(1-\rho-\lambda)}{\rho k}I_{k})))=0
    det(\frac{J_{k}}{k}-(\frac{-(1-\rho-\lambda)}{\rho k}I_{k}))=0

    Now this is merely the equation for determining the eigenvalues of \frac{J_{k}}{k}. Since, \frac{J_{k}}{k} is idempotent the eigenvalues of \frac{J_{k}}{k} must be either zero or one. In fact, since this matrix has rank one, \frac{J_{k}}{k} has eigenvalues one with multiplicity one and zero with multiplicity (k-1). Therefore the eigenvalues of \Sigma can be found by setting \frac{-(1-\rho-\lambda)}{\rho k}=0 and \frac{-(1-\rho-\lambda)}{\rho k}=1. This yields \lambda=1+(k-1)\rho with multiplicity 1 and (1-\rho) with multiplicity (k-1) which are exactly the eigenvalues of \Sigma.

    Cheers.

  • More H1N1 (in the wild)

    A good excerpt from this article, By Gary Kreps, Ph.D, and Rebecca Goldin, Ph.D, November 17, 2009:

    “Unlike the seasonal flu, H1N1 frequently attacks children. The CDC calculates that 179 flu-related pediatric deaths have occurred in the U.S since last April. Of these, one was due to the seasonal flu and 156 were due to H1N1. (The other 22 were due to a Type A influenza with an unidentified sub-type.) Compare those figures to the 2006-2007 flu season, when only 68 total pediatric deaths were linked to the seasonal flu. Thus, H1N1 has killed almost twice as many kids in the first month of this year’s flu season as the seasonal flu killed in an entire year during 2006-2007. The stakes are high for pregnant women as well, who constitute about one percent of the population but six percent of the deaths attributed to H1N1.

    The media could help parents sort this out by framing this story in terms of comparative risk. Some parents may be willing throw the dice, reasoning that the absolute risk to their children is low. Instead they should compare the risk with that of other viruses for which vaccinations are now standard. Chicken pox, which used to take kids out of school for one or two weeks, was widespread until a vaccine became available in 1995. Before the vaccine, 100 to 150 people died each year from the disease, and more than 10,000 were hospitalized. This cost to society was considered high enough that 46 states now require children to get vaccinated in order to attend school.

    As this comparison makes clear, the decision to vaccinate against H1N1 should be a slam dunk. The danger may not be apocalyptic, but it is very real. Unless America’s parents get this message, it is their children who will suffer most from confusion and misinformation. In fact, once you get past the conspiracy theories and myths, the development of the H1N1 vaccine is a genuine success story of government and industry working together to serve the public interest. But it’s being undermined by a failure to get the real story out to the very people whose lives may depend upon it.”

    Cheers.

  • Short Term Global Warming (in the wild)

    Let me start by saying I’m not an expert on global warming. I’m absolutely sure the earth is getting warmer (think melting ice caps) and very sure that it is caused by humans (green house gases). But who knows. Remember, if someone yells their dissenting opinion loud enough, it becomes fact, right?

    Anyway, I’ve read some articles about how global warming has “stopped” in the last ten years. For instance, this article: “Climatologists Baffled by Global Warming Time-Out” states: “At present, however, the warming is taking a break,” confirms meteorologist Mojib Latif of the Leibniz Institute of Marine Sciences in the northern German city of Kiel. Latif, one of Germany’s best-known climatologists, says that the temperature curve has reached a plateau. “There can be no argument about that,” he says. “We have to face that fact.”

    It goes on to say: “Even though the temperature standstill probably has no effect on the long-term warming trend, it does raise doubts about the predictive value of climate models, and it is also a political issue. For months, climate change skeptics have been gloating over the findings on their Internet forums. This has prompted many a climatologist to treat the temperature data in public with a sense of shame, thereby damaging their own credibility.”

    This sounds like he is claiming that the warming has stopped. I disagree with this. You can have a system that is, on the average increasing over the long term, while still observing very flat or even declining trends when we know the overall system is increasing. That doesn’t mean that the system isn’t increasing, it just means we’ve seen one realization of the random system that hasn’t increased entirely by chance.

    Here is a simulation experiment. (All these numbers are made up, but they prove the point):
    Consider at year 1 the average temperature is 75 degrees. Call this x[1]. Then at year two we observe a realization from a normal random variable whose mean is 1.005*75 with standard deviation 1. Call this x[2]. x[3], the temperature in the the third year, will than be an observation from a normal random variable with mean 1.005*x[2] and standard deviation 1.

    Over the long run this is an increasing sequence, but let’s look at what happens in the relatively short term. I simulated 10,000 of these such chains for 100 “years” each.

    After 5 years 20.4% of these sequences we below the starting temperature of 75 degrees. After 10 years, 12.8% were below 75 degrees. Think about that. We have a known increasing sequence and after 10 years, 12.8% of them ended below where they started. Global warming is like this. We can see small periods of decline, in fact we EXPECT to see small periods of decline, within this increasing sequence.

    What happens when we look at this sequence after 50 years? 0.006% are below the starting temperature of 75 degrees. After 100 years, 0 are below 75 degrees.

    So to say that global warming is “taking a break” based on ten years of evidence seems like bad science to me. And this is certainly not evidence invalidating the long term usefulness of climate change models.

    Cheers.

  • Obesity and Diabetes (in the wild)

    Estimated County-Level Prevalence of Diabetes and Obesity — United States, 2007.

    Also be sure to check out this very nice figure which shows the United States at the county level by rate of diabetes and obesity. It’s really striking in this figure just how centralized high levels of obesity are in the southeast and Appalachia. It also show the close relationship between high levels of obesity with a large prevalence of diabetes.

    There are two other pockets that strike me as interesting: Northeastern Arizona and the border of North and South Dakota. What explains those high levels of obesity and diabetes? The first thing that comes to mind is that those could possibly be areas with a large population of Native Americans. So I Googled “Native American’s and obesity” and I found this study: “The epidemic of obesity in American Indian communities and the need for childhood obesity-prevention programs.” The first few sentences of the abstract are: “American Indians of all ages and both sexes have a high prevalence of obesity. The high prevalence of diabetes mellitus in American Indians shows the adverse effects that obesity has in these communities. Obesity has become a major health problem in American Indians only in the past 1–2 generations and is believed to be associated with the relative abundance of high-fat foods and the rapid change from active to sedentary lifestyles. Intervention studies are urgently needed in American Indian communities to develop and test effective strategies for weight reduction.”

    And this study: “Prevalence of Obesity Among US Preschool Children in Different Racial and Ethnic Groups” (Sarah E. Anderson, PhD; Robert C. Whitaker, MD, MPH. Arch Pediatr Adolesc Med. 2009;163(4):344-348.) which claims these results: “Results Obesity prevalence among 4-year-old US children (mean age, 52.3 months) was 18.4% (95% confidence interval [CI], 17.1%-19.8%). Obesity prevalence differed by racial/ethnic group (P < .001): American Indian/Native Alaskan, 31.2% (95% CI, 24.6%-37.8%); Hispanic, 22.0% (95% CI, 19.5%-24.5%); non-Hispanic black, 20.8% (95% CI, 17.8%-23.7%); non-Hispanic white, 15.9% (95% CI, 14.3%-17.5%); and Asian, 12.8% (95% CI, 10.0%-15.6%). All pairwise differences in obesity prevalence between racial/ethnic groups were statistically significant after a Bonferroni adjustment (P < .005) except for those between Hispanic and non-Hispanic black children and between non-Hispanic white and Asian children."

    Cheers.

  • Stock Market (in the wild)

    This article is about Andrew Patton from Duke (which was tweeted by NISSSAMSI had this interesting tidbit about the stock market in it:

    “‘Unfortunately, stock prices are almost impossible to predict,’ Patton said. ‘But what we can look at are things like risk and correlation.

    ‘For instance, it is well known that a given bundle of stocks often decline in value together, but those very same stocks rarely increase in value together,’ he said. ‘In other words, there’s something very different going on in a bear market than in a bull market, and my research tries to capture that difference.’”

    I’m pretty sure that is really interesting.

    Cheers.

  • Pretty pictures from R (in the wild)

    Awesome: The R graphics gallery.

    Cheers.

  • US vs the world: Higher Education (in the wild)

    My friend emailed me this article this afternoon:Apples and Oranges in higher education.

    It’s all about the use and misuse of statistics in comparing the United State higher education system to other countries. Pretty interesting stuff.

    Cheers.