31 August, 2005

Long hours on statistics...

I am, without question, an American. Why do I say this? I spent ten and a half hours at work yesterday (nine and a half, if you take away the hour-long bicycle commute). It's true that I spent only eight hours at work the day before (seven), but I did a bit of work at home, too. It's been exhausting, and if any of my readers are left, thanks for sticking around...

I actually have a lot to write about; I just don't have the time to write it. Without question, most of my time has been spent on preparing for statistics, and grading homework quizzes. I hope it will please Alessandra to learn that I've decided to print one of the articles she's linked to, and use it as an exercise in identifying problems with statistical citations. I actually like the article very much, but I think it has a few problems that the book identifies. Tomorrow's homework will be:

  1. a biased sample ("man-on-the-street", self-selected, etc.);
  2. problems with experimental or observational studies (unnatural setting, Hawthorne effect, confounding variable, ethics, unclear cause-and-effect, reliance on others' data);
  3. misuse of statistics (suspect sample, no control group, ambiguous average, changing the subject, detached statistics, implied connection, bad question).
I think the hardest one for the students will be identifying a biased sample, but the biased sample I have in mind is the author herself, who uses her own life as an example. The point is that what makes for excellent writing also makes for bad statistics.

I also picked an article from the Washington Post on problems with measuring the economy.

As for trying to get at the students' interest: I've taken another professor's group projects and issued it to the group. Hopefully they will use their own interests and experience to choose a survey topic and the questions that follow. In any case, they should be involved in the project, which makes for better learning than listening to me.

I'd like to think that there are some good signs. Homework scores aren't so good, but students are disagreeing with me in class about whether an answer is good or bad, and their arguments are often on good grounds. Some of these answers aren't very clear-cut (the other professor has told me that statisticians disagree on them) so there's wiggle room.

Yesterday, on the other hand, I cited Nazi medical experiments as an example of an unethical experimental study. A student suggested the Tuskegee Project; I initially agreed, but then the student went on to say that the government gave the subjects syphilis, much as Dr. Mengele injected his subjects with poisons (or whatever he did). That struck me as inaccurate, and subsequent research seems to confirm my understanding: the doctors didn't give the subjects syphilis, but deliberately failed to tell the patients that they had syphilis. Instead, they told the patients that they had "bad blood".

Does anyone know otherwise? Here's a Wikipedia reference; nothing suggests that the men were deliberately infected with syphilis.


Alessandra said...

"it will please Alessandra to learn that I've decided to print..."

lol :-) Hey, glad it could be of use! Not that I know stats in any depth, but I do know what a circus the news media (and other venues) are when it comes to presenting information.

Although the numbers from the data collected may have problems (or how they were presented), I actually didnt analyze them in depth, I don't doubt that main point that the article states, cohabitation can adversely affect the plan of a posterior marriage in a sizable number of couples.

Someone once recommended a book to me, exactly on this subject of manipulating numbers, and if I remember correctly, it was written by a mathematician. I didn't have time to read it at the time, but I know it has really good examples similar to those you cite and it would be good for your class. The book is exclusively about ways that numbers are cited that give an accurate impression, while in truth, it's a very distorted presentation. Unfortunately, my mind is drawing a blank on title/author.

Glad to know you are surviving your first week! And as for the amount of work... well, it could be worse. :-)

I am also glad you have been able to find a few minutes to "pop in" your blog.

Re the Tusk. experiment, it's been too long since I had it described to me, and I don't remember the details anymore. I'm sure a search could easily yield some reliable summaries.

And re your students discussing with you about what makes a good /bad answer and that you thought they have good reasoning to base their questionings on, that is a very good sign. :-) I hope you will have fun, even with your favorite subject course!! :-)

jack perry said...

Well, I guess I have to apologize now; I changed my mind today about that. They have enough homework for now as it is.

On the other hand, I conducted surveys in class and used the results to make frequency distributions and histograms. For one class I asked, "How many drinks of soda did you have last night?" For the other class I asked, "How many credit hours are you taking?"

Most of them were taking 12-15. A handful took more, and one claimed to be taking 21. The class gasped when they heard that.

What surprised me the most, however, was that some of them were taking 12.5 or 16.5 credits. I didn't know that was even possible, but they explained that there some exercise science classes are worth only 1.5 credits or so.