Final Blog Posts

Hello everyone! We had a lot of fun didn't we.

I'm happy I learned a lot in this class, and I am now moving on to an actually statistics 145 class. I wish I had done this before this class though. I didn't know it would be so hard with the math, and I really struggled through it. I wasn't the only one I'm sure. But overall I had fun.

The medical analysis was my favorite part, I think. That's why I wanted to do that as a part of the final project.

Final Thoughts

Whew! The semester is almost over! I hope everyone does well on all of their finals, and hope to see you guys around campus next fall. This class has been most helpful very instructive in helping me learn more about statistics but to be more critical of scientific papers. As a budding data collection junkie, my questions to the class are these: What did you find most interesting in the class? What did you find most helpful? (This is not a scientific poll).

-Lee

White collar criminals

                After reading the article from the Political Economy Research Institute I began to think about the cost of this sort of publication on public health and welfare.  While we are often able to focus on the human cost of a homicide (as we saw earlier in the semester in the Philidelphia homicide data) it is difficult to think about the widespread cost of white collar crime.  The U.S. Department of Justice quantified the monetary cost of white collar crime at anywhere from 400 billion to 1.7 trillion dollars.  Being that I cannot adequately express what that amount of money may do to a human being I thought about a much smaller example, comparatively so, of white collar crime in recent years.  Bernie Madoff’s ponzi scheme cost his investors over 64 billion dollars and financially crippled more than 2,400 (the other half of Madoff’s clientele are reported to have not been injured financially).  Add to this minor financial issue that his actions were the estimated cause of three different suicides, including his own son, and the toll seems worse still.  As a result of the Grecian austerity measures there were a total of five homicides during the riots and the suicide rate has doubled prior to before the announcement of the measures.  I suppose that if we were to directly attribute just these five homicides indirectly to Rogoff and Reinhart then we could say that they were only the catalyst for 2.5 deaths each and that doesn’t seem quite as bad as Madoff’s crime.  Does it?

http://www.peri.umass.edu/fileadmin/pdf/working_papers/working_papers_301-350/WP322.pdf

http://en.wikipedia.org/wiki/Madoff_investment_scandal#Suicides

https://www.ncjrs.gov/App/publications/Abstract.aspx?id=167026

Doesn't matter how you text, it's all equally distracting while driving

A study was reported on ABC News today that found no difference in the level of distraction caused by manual texting or voice-to-text messaging. The study was conducted by the Texas A&M Transportation Institute, and funded by the US Dept. of Transportation Research and Innovative Technology Administration.
The researchers took 43 research participants and put them on a closed course to drive an actual vehicle, a method which, according to ABC News, has never been done before. Each participant drove the course four times: the first time without any texting, once with Siri voice-to-text, once with Android Vlingo voice-to-text, and finally while texting manually. The researchers observed how long each texting method took, and also how quickly the drivers responded to randomly placed lights on the course.
The researchers found that each texting method delayed the drivers' reaction times, all about double what their reaction time was in the control.
What I found most surprising is that the drivers spent "significant" amounts of time not looking at the road regardless of which texting method they were using. I suppose that looking up at the road while composing a text out loud isn't the same as really watching the road, but I still find it surprising enough to hope for further study and replication. Another unexpected result was that the voice-to-text texters took longer to complete their message than manual texters. I would have thought that speaking a message out loud would be much quicker than typing it, but maybe the participants also had to set up the voice-to-text application and send the message to the right person.
I also wonder what the participants were asked to text--if they were assigned a message, in order to standardize the lengths of the messages, or if they were allowed to send anything they liked. I would guess that it takes more attention trying to dictate or type words that are not your own, but I couldn't say how significant that difference is.
However, of course the most suspect part of this study is its funding source and timing, being published during National Distracted Driving Awareness Month, at the same time that the Texas DOT is sponsoring anti-distracted driving promotion and advertisement.
Thankfully, the ABC News article reported this study well, including the timing and funding issues, and also included the study manager's statement that “Understanding the distracted driving issue is an evolving process, and this study is but one step in that process...We believe it’s a useful step, and we’re eager to see what other studies may find.”
The inclusion of this statement, saying that this study is just one step of many to be taken, and not the source of every right answer we've ever been looking for, is refreshing after so many overzealous media blowups about studies similar to this one, that have counterintuitive and/or surprising conclusions and real-life effects on the safety of many Americans.

http://www.abcnews4.com/story/22059954/voice-to-text-apps-offer-no-driving-safety-benefit-as-with-manual-texting-reaction-times-double-according-to-texas-am-transportation-institute

Statistics

I’ve been doing a lot of research lately and I found that one of the easiest places to find statistics is through the CDC website. I was researching some information about Fetal Alcohol Syndrome and was easily able to find statistics on this syndrome throughout the entire United States. After doing the research for my presentation, I found this website to be a very easy resource for the public to use (something we need more of). Many people have trouble finding data about certain topics, but I found that the CDC made it very easy for the public to see what is going on within society. Also, because the education levels within the U.S. are continuing to decrease, I found it helpful that the CDC made it very easy to understand the statistics they were presenting. Overall, I wish there were more websites that made their statistics more accessible, as well as easy to understand. This could help with the problem of people believing everything they see in research and on the media. You can see how easy the website is from the link below:

http://www.cdc.gov/NCBDDD/fasd/data.html

How Circumcision may Reduce HIV

Circumcision reduces the risk of a man both transmitting or acquiring H.I.V., and other sexually transmitted diseases. However, until now, we have not known why this is. Researchers in Uganda studied 79 men randomly assigned to circumcision and 77 men who remained uncircumcised. After a year, the amount of bacteria decreased in both groups (may be due to behavioral changes caused by the study). However, the decreases were much larger in those of the circumcised sample. They also found that the diversity of bacteria among those who were circumcised decreased. I found this article interesting because they are trying to find ways to decrease one’s risk of contracting H.I.V. From this study, they believe that bacteria may play a role in attracting T-cells or cause inflammation that makes the transmission of H.I.V. more likely.

http://well.blogs.nytimes.com/2013/04/18/how-circumcision-may-stem-h-i-v/?ref=health

A Recent Critique of the Analysis Behind Austerity Policies

You may have seen some headlines recently about a mistake in an Excel spreadsheet being the driver of austerity economic policies of the past few years (such as this article in the Economist).

I encourage you to look at this paper: "Does High Public Debt Consistently Stifle Economic Growth? A Critique of Reinhart and Rogo ff" by Herndon, Ash and Pollin. Follow this link and click on "Download the paper here."

For you this paper is interesting because it is a good example of a critique of another researcher's analysis, much like the paper that I am asking you to write. They begin with a brief synopsis of the key points in the original paper, provide a whole section discussing the important practical implications of the paper, and then show their own re-analysis of the data tying it back to the practical implications. Your own paper should have a similar structure, but it will look rather different. First of all, the introduction of the Herndon, Ash & Pollin (HAP) paper is written assuming the reader is fairly familiar with the Reinhart & Rogoff (RR) paper as well as non-parametric statistics. Also, they have far more statistical re-analysis than I would expect in your paper.

HAP includes three main criticisms of RR. First, there is some omission of data in the calculations. Second, the analysis includes questionable assumptions. Third, HAP questions the interpretation that the relationship between debt ratio and GDP growth changes drastically at the 90% debt ratio point. Any one of these criticisms would be plenty for your own paper. HAP provides re-calculations and plots new graphs to illustrate the re-interpretation, as you should too.

Thinking about this subject in general (besides being a good paper example), both the HAP and RR papers are just two pieces in an ongoing dialog about the analysis and interpretation of this data. If you notice, the HAP paper includes not only the written PDF of the paper, but also the data, code, and explanations of the various files. Furthermore, they publish updates to the paper as corrections and clarifications arise. This is not quite a dialog yet, but it does reflect the idea that their work is never the "final" word.

This blog post by a research scientist at Berkeley discusses how a true dialog that includes analysts, journalists and policy makers could take place surrounding such data and analysis. Fernando Perez writes:

It's absolutely unacceptable that major policy decisions that impact millions worldwide had until now hinged effectively on the unverified word of two scientists: no matter how competent or honorable they may be, we know everybody makes mistakes, and in this case there were both egregious errors and debatable assumptions.

The Beauty of Data Visualization

Earlier in the semester we talked about how statistics are visually represented in print and news media, and the problem of sacrificing content for the sake of creating something that is simply visually pleasing. I love TED Talks (Technology, Education, Design) and if you don't know them you should. I found this really nice one from TEDGlobal 2010 on "The Beauty of Data Visualization." It presents a starkly different view point from the one we discussed in class, suggesting that creating visuals for data can often enhance the statistics and illuminate unseen connections instead of dumbing-down or botching the stats.

The Road to Fraud...

In doing research for our group project I found this nice article on "The Road to Fraud" from 2011 on Nature.com. The central point the author makes is that "To understand fraud, we should think about how it begins and escalates, not how it ends. By the time such fraud is exposed, bad choices that would usually lead to only minor transgressions have escalated into outright career-killing behavior." I think this is important, and perhaps the answer I was looking for in regards to the question of avoiding the "slippery slope" or "trap." It would be much more productive to be actively checking the ethics of one's research along the way and rectify any problems as opposed to looking back and simply trying to apologize for bad science after the damage is already done. Another important point she makes is 

But a heavy focus on fraudsters may also conveniently divert our attention from the fraudster within us all. Who cannot find places where they took a first step, or perhaps several steps, down one slippery slope or another? The road to fraud probably starts out with a step taken because of some egotistical fear or anxiety — fear of losing someone's respect, for example, or of letting others down, the fear of being seen as a loser, of being a failure, or of not getting the job, the grant or the award that one covets. 

In many ways, I think this statement goes back to the discussion we had in class about what to do with smaller more "insignificant" studies as I think we all agreed that the pressure to publish big, never-before-seen discoveries often leads to fraud. Thus it's important to remember that scientists are not infallible and genuine mistakes are made, and should be identified as such and corrected before they are let slide to the point of full-blown fraud due to fear of shame or ridicule from the scientific community.  

SHC/SUB might as well take advantage of hearing loss increases

After having to go on an adventure trying to find hearing aid batteries today (which were nowhere to be found on campus), I decided to look up hearing loss statistics for young Americans.
According to Allen Audiology, 20% of all Americans experience some kind of hearing loss. What was a surprise to me was its prevalence in veterans: 

"Of the over 800,000 veterans who received disability benefits that year, 148,000 (18.5%) received them for tinnitus or hearing loss; by comparison, the number receiving compensation for post-traumatic stress disorder (PTSD) was 42,700 (5.3%)." 

I immediately started wondering about how many actual cases of PTSD go unreported and/or untreated, versus hearing damage, which has basically zero stigma for soldiers, and can be diagnosed quantitatively. That being said, it does make sense that so many veterans would have hearing damage, because as the article goes on to say, a soldier's instinct when in combat, fighting among all kinds of firepower, isn't to put in their ear plugs.

But going back to college campuses--you don't need to be in combat to develop hearing loss. According to the American Medical Association, reported in a Daily Mail article, hearing loss has increased 30% in the past 20 years in teenagers, thanks in large part to being voluntarily blasted with loud music through headphones and earbuds. The European Commission recommended a 60/60 guideline, meaning that people should listen to music for no longer than 60 minutes at a time, at a maximum of 60% volume. However, a study done by the Royal National Institute for Deaf People found that 66% of MP3 player users go beyond that recommendation, putting their hearing at significant risk.

Since hearing loss is still incurable for people of most incomes, and there are so many young students and veterans here at UNM, I would hope that the SHC and the SUB would use these statistics to start making profits off of hearing aid batteries, so I don't have to go all the way to Walgreen's on my next emergency pickup.

http://www.allenaudiology.com/military-service-resulting-in-hearing-loss-what-should-we-do/

http://www.pamf.org/teen/health/diseases/mp3hearing.html

http://www.dailymail.co.uk/health/article-1303998/iPods-damaging-childrens-ears-Teenage-hearing-loss-rises-30-20-years.html

Poor North Dakota

This is just a funny collection of statistics from Pleated-Jeans, making the United States of Shame. It's a graphic showing what each state is ranked worst at from various statistics. Some of them sound real and based on quantifiable numbers, like West Virginia being the worst place for heart attacks, while some are more dubious, like the "Ugliest Residents" in North Dakota.
It's just an example of how actual statistics (the sources for each state are below the chart) can be used more casually by online entertainers and dispensed to the public.

http://www.pleated-jeans.com/2011/01/24/the-united-states-of-shame-chart/

Statistical Literacy

At the end of Joel Best's article, Lies, Calculations and Constructions: Beyond How to Lie with Statistics, he said that someone must take the initiative in educating people in statistics.  He also said he doubts it will be Ph.D. statisticians who will do this because they might feel it is beneath them.  Today we talked about the about free online classes that are available to anyone and the pros and cons of them.  Although diversity in teaching methods will be virtually nonexistent in these classes, at least someone is taking the initiative.  The general public could be able to improve their statistical literacy, and if it leads to more interest in the subject, there could be a desire to take courses at a community college or university.  Classes like this are able to help the public understand statistics and not be fooled by the media's misinterpretation.

Skeptic, not suspicious

I think the biggest thing I’ve learned from this class is just to be skeptical of mostly anything that is reported to me. The studies and statistics headlines that I glance at on my Yahoo! homepage every day are all chosen by people trained in marketing, written by people trained in communications or journalism, about studies done by scientists that did not conduct infallible research. Even if the original study is sound, it’s still like a game of telephone—the conclusion goes through several people, each with different objectives, before the information finally reaches the masses, to be received by people of vastly different educational and belief backgrounds.

A complaint that I’m sure you’ve heard before is that wine/chocolate/coffee/ is good for you, but it was bad for you three months ago, and good for you again six months ago, and so on. This complaint is made because people often don’t realize that every research study is just that—research, and not a claim to be the absolute truth. I doubt that people consciously think that, but when a study is reported on the news that a group of scientists with PhD’s conducted, it seems as though enough people trust it to be true for it to really be true. 

So unfortunately, the effect of the constantly “changing” reports that recommend or discourage this food or that activity has not been “OK, so none of the studies I hear about are the Answer, but they’re still good investigations,” but instead, “I don’t know about those scientists, they can’t agree on anything, I better Google it instead.” There’s a difference between being skeptical consumer of statistics, and being suspicious of scientists, which is a line that many people still have yet to cross.

To Observe, To Question, to Understand

To Observe, To Question, to Understand

I had taken this course this semester to learn about how statistics are utilized in social, economic and political research that we come across everyday within the media as well as scientific studies. I left my motif to that...until we began class.

I am one of those people that are "afraid" of numbers. I always have been. They were a driving force to take me further away from the fields I was (and still am) interested in (Astronomy, Engineering...). They illustrate a principle of conveyed antagonism, etched within the confines of algorithms and notorious numerical variations. They are, as you could say-My oppressor, and I was the oppressed, sanctioned off from a parallel universe of astronomical computations and gaussian beams. Not only this, but the research necessary to compile enough notes and data to make since of it all is riddled with wondrous statistical analysis and process. However, I recognize now that this mysterious (often fatal...for me anyways), voice with numbers-specifically statistics-is not so hard to comprehend if you take the time to observe, to question and then to understand. Statistics plays such an important role. And humans are in charge of them. On a day-to-day basis, we (the "consumer"), are swarmed with statistics, who need strong adversaries. In this class, that is what we have learned to be. It has made me a stronger person, and I am sure it has affected many of you as well. I question everything now-every number, every correlation and every stated causation said to be fact. As we have discussed today, it is crucial to require this same statistical analysis from everyone who is a presumed "Producer," "Purveyor," "Consumer," or "Decomposer."

As a community, society, and an entire nation, actually understanding where our news is coming from, and pushing people into teaching them how to question, should be a top priority. That is where scientific breakthrough comes from, and that is where cooperation is made. We are shown from a young age that math (the vast majority of it), is scary.As shown by the National Center for Education Statistics, the percentage of eighth-graders taking algred went from 16% to 31% from 1990 to 2007. California had 4% of their eighth-graders taking algebra in 2009 (USAToday). This could be seen as a good thing, but also a bad thing. Many state that algebra is so abstract, that by the time kids are in middle and high school, most kids have difficulty in really understanding the math material given because basic math was gone over rather quickly when they were young. Overall, this may be detrimental to our society. Why do I bring this up? Because overall, the level of our education system represents the level of educational values that our officials are fine with. We should always desire progress, but with budget cuts to schools and headstart programs, to summer camps and financial aid, it is difficult to see how there could ever be a surplus of mathematician and engineer dreamers in our school system. This class has at least shown me, that math is useful, possible and interesting. And that is how our educational system should treat math. I feel that if this changes through each school grade, then we may also see changes in not only future math and science research, but also changes in media portrayals and use of statistics. It can change politics, the economy, and ultimately, society. And that is what I hope for. 

http://usatoday30.usatoday.com/news/opinion/forum/story/2012-07-09/math-education-remedial-algebra/56118128/1

Also, for some satire-Report: Chinese Third-Graders Falling Behind U.S. High School Students in Math, Science

By the most reputable source on the internet-The Onion

http://www.theonion.com/articles/report-chinese-thirdgraders-falling-behind-us-high,31464/

 

In reflecting back upon the class I've been thinking about the idea of "lying with statistics" and especially with Dr. Potti's case that we talked about recently. I think, and I'm sure I'm not the only one, that I'm probably still a guilty of holding a little of that blind faith in numbers that Gould talked about at the very beginning of his book. I think with all the unknowns of the world we'd like to believe that numbers stand as something concrete and definite with absolute meaning. And the numbers themselves do, but we forget that there are people behind those numbers and that's where things get messy.

In the case of Dr. Potti the numbers had the potential to do great things, but were horribly misused at a great cost. In the original article published in 2006 on Dr. Potti's research notes the power that his discoveries held: "The test could improve the prospects of patients who currently suffer the side effects of chemotherapies without knowing if they are undergoing a treatment that has a good chance of working for them." Obviously, we all know how this turned out. It is The Economist's later article from 2011 that I find particularly striking for it notes the many complications that could have contributed to the Duke scandal apart from simply blaming Dr. Potti. I think the challenges the point out in regards to peer reviewing are especially interesting: 

the process of peer review relies (as it always has done) on the goodwill of workers in the field, who have jobs of their own and frequently cannot spend the time needed to check other people's papers in a suitably thorough manner [...] Moreover, the methods sections of papers are supposed to provide enough information for others to replicate an experiment, but often do not. Dodgy work will out eventually, as it is found not to fit in with other, more reliable discoveries. But that all takes time and money.

Thus we can see how it takes a whole team of individuals to produce the sort of numbers that can be replicated and that people can rely on. This is why it's so important to remember that numbers are indeed fallible, but not necessarily out of malicious intent, but because of the extreme amount of effort it takes to create the sort of exemplary research that "consumers" of statistics or the general public desire. Overall the article helped put some of the Duke case in perspective for me––I'm absolutely not excusing Dr. Potti–– but it simply helped to provide some more background information.

But Where IS the Data?

One of things we have stressed in class lately is the importance of peer viewing other's papers. We have spoke to its benefits and have been able to find many problems with data analysis with only using simple statistical methods. Mistakes and cover ups appear to be quite common in published papers.

The problem is finding the raw data to redo analysis. Most papers published, on the web at least, are in their polished and concise form and do not contain data files. We simply get a nice and neat summary table that highlights the results that ultimately support the researchers' hypothesis. While this is not inherently bad and is VERY pervasive, it does limit other people's chance to check your work.

The transparency we crave is lost, and the more cynical readers will start to question your results and potentially your motives for not providing the data. That being said, most good researchers will gladly provide one with the necessary data files if asked, but this may take longer than one has time for.

What I propose is requiring data files, or at least links to those data files, to be readily available to the viewer. We are in an information age, let's live up to those expectations.

Only One Way to Deal with a Trap

In the words of Admiral Ackbar, “It's a trap!” Unlike Admiral Ackbar our fleet of scientific research is under the attack of misconduct which can only be solved with a "Deathstarring." There are several ways to approach the issue at hand, either we take the enlightened path, as noted in one of AJ's previous posts, or some other less optimistic route. Unfortunately humanity is good at botching any number of things, from science to such things as religious injustice. While I am no savior from these follies, I do believe that one such possible solution takes us down a more conservative route. I propose the implementation of stricter punishment for individuals whom knowingly commit fraud in science. While others have pointed out that the suing of individuals on a case by case basis does occur, usually these people receive little in punishment (Potti), less their own defamation in scientific circles. The main outcome of the implementation of stricter punishment would be to deter individuals from conducting distorted science. As others believe that transparency will be our enlightenment, pessimistically I do not believed we have reached the societal capability of such great bounds, yet. A more effective means to solving the problem under our current societal structure would be in the form of a swift, firm slap to those who wantonly debilitate science. Earlier today, someone mentioned our mother-child type relationship with science with which I agree. I also agree that when there is intentional wrong doing, some people deserve a thorough “time out.”

Misconduct accounts for the majority of retracted scientific publications

After a detailed review of all 2,047 biomedical and life-science research articles, Ferric C. Fang found that only 21.3% of retractions were due to error. However, 67.4% of retractions were due to misconduct. The following are the percentages for each type of misconduct:

-Fraud or suspected fraud (43.4%)
-duplicated publication (14.2%)
-plagiarism (9.8%)

Since 1975, the amount of papers retracted due to fraud has increased 10-fold. I found it interesting that the real problem is not that people have errors in their statistics, but that they just act unethically. It makes me feel like this society has a bigger problem in regards to research than we originally thought.

http://www.pnas.org/content/early/2012/09/27/1212247109.abstract

Surprising Results

As I was doing some research for the upcoming presentation, I came across an article that presents research that links being overweight to a lower risk of mortality. I found these results interesting because I have always been told the opposite. The conclusion presented in the Journal of the American Medical Association stated that the detailed review of over 100 previously published papers connected body weight and mortality risk among 2.88 million study participants around the world. Although the new research confirms that obese people tend to die before those of normal people, the findings also suggest that people who are overweight (but not overweight) may live longer than those who have a clinically normal weight. Although these results have gotten a lot of negative feedback, dozens of other researchers have been able to reproduce the same results. Overall, I found these results to be interesting.

http://www.cnn.com/2013/01/02/health/overweight-mortality

Yale Autism Seminar

Although we haven't been talking about Autism all that much lately, I know that many people expressed general questions about it when we were discussing the Lancet Papers and Vaccine War. One question that seemed to be tossed around frequently was in regards to how it is specifically diagnosed in children, and I found this wonderful (free!) web series from the Yale Autism Seminar which has 14 videos related to Autism Spectrum Disorders. I haven't encountered any mention of the vaccine-Autism link yet but I also haven't watched all of them. Just thought this might be helpful for anyone who was really interested in the topic or might be doing their project on something Autism related.