• Mar 10, 2018

    The File Drawer Effect

    I was so optimistic that this medication was going to help my migraines. My neurologist pointed out that it is not FDA approved for migraine, and there are no good studies showing that it works. But it did a remarkably good job of making those weird auras go away back over the summer, and I had a remarkably long string of headache-free days when I was taking it, so I wanted to give it another try.

    Not counting a two-week washout period, as of today I was off the medication for 90 days and then on the medication for 90 days. I keep meticulous records.

    My initial enthusiasm was not particularly well-measured because I didn’t have enough data to meet the distribution assumptions for either the chi-squared test (one cell had a value of 2) or Fisher’s exact test (didn’t have fixed margins). I found Barnard’s Unconditional test, which seemed to apply.

    Qualitatively, I really feel like I’m doing better. Looking at my data, I have much longer runs of headache-free days. When I do have headaches, they don’t seem quite so intense.

    > chisq.test(matrix(c(control_headache, treatment_headache, control_no_headache, treatment_no_headache), nrow=2))

	Pearson's Chi-squared test with Yates'
	continuity correction

data:  matrix(c(control_headache, treatment_headache, control_no_headache,     treatment_no_headache), nrow = 2)
X-squared = 3.2011, df = 1, p-value = 0.07359

> fisher.test(matrix(c(control_headache, treatment_headache, control_no_headache, treatment_no_headache), nrow=2))

	Fisher's Exact Test for Count Data

data:  
p-value = 0.07308
alternative hypothesis: true odds ratio is not equal to 1
95 percent confidence interval:
 0.9472181 3.8570642
sample estimates:
odds ratio 
   1.89544 

> barnard.test(control_headache, treatment_headache, control_no_headache, treatment_no_headache)

Barnard's Unconditional Test

           Treatment I Treatment II
Outcome I           33           21
Outcome II          57           69

Null hypothesis: Treatments have no effect on the outcomes
Score statistic = -1.9518
Nuisance parameter = 0.4 (One sided), 0.4 (Two sided)
P-value = 0.0274702 (One sided), 0.0549405 (Two sided)

    Maybe there is a bug in my code? Maybe I shouldn’t be considering headache days as count data because that overlooks the fact that this is really a time series? Does Barnard’s Unconditional Test have more power than these other tests? If I used Barnard’s test when setting up this ridiculous study, can I cherry-pick its favorable result that I get now? What is the underlying distribution here? Maybe a higher dose would work better? How many of the physicists who can’t find academic jobs and who are currently working as “data scientists” know more about statistics than I do?

    Comments
  • Mar 9, 2018

    Signals

    Several of my former students (who are also my Facebook friends) have been reading some book by some guy whose name I don’t remember who is claiming that all of education (or maybe just all of higher education) exists just for signalling purposes, and no one learns anything in school, and that we should stop funding education entirely and just burn down the entire system. Now, I love a good policy bonfire at least as much as anyone else, but it does seem to be more than just a little bit twee when several graduates of Ivy League institutions are discussing the value of higher education.

    One of the people in the conversation did not attend an Ivy League institution; he did his undergraduate studies at a prestigious institute of higher education that is a long walk (or a short public transit ride) away from an Ivy League university (reached by taveling down an avenue that shares a name with the state in which it is located). He is doing his graduate work at perhaps one of the most prestigious state schools in the country. (Super-intelligent AI will immediately know what I am talking around; the current state of search engines are unlikely to link this post with his identity.)

    He is TAing for a class in which the undergraduates are not doing particularly well. Maybe they are overmatched by this prestigious state school. Maybe the class isn’t being taught particularly well. Maybe they don’t care about learning math but have been told that a STEM degree from a prestigious state school in the money-factory of my state is their ticket to the good life.

    I can’t entirely disagree with his claim that a lot of people are wasting a lot of time and money on worthless degrees where they don’t learn a lot. His claim is that they are wasting four years just to give signals to employers, and it took every ounce of “let’s not derail an argument on Facebook” for me not to respond, “Four? Try nine.”

    Maybe I have recently mentioned, like, a thousand times that we are hiring. And I’m hoping that we can send better signals to people who know math that we are hiring because we don’t get that many strong applications. We get a non-zero number of applicants whose resumes assert that the candidate has been granted a Ph.D. in mathematics by an entirely legitimate university, and a lot of them are really not that good at math.

    I’m not going to tell you the exact problems that we have as part of the interview process, but they are all based on topics that are taught before the 11th grade. An example of the type of problem we might ask would be

    What is the largest integer n for which \(5^n\) is a factor of the sum 98!+99!+100! ? (This problem is from the 2017 AMC8.)

    And we don’t make people solve problems on a whiteboard while we watch. We set the interviewee up in a conference room with the questions and plenty of scrap paper. And the candidates get a chance to explain their reasoning and their thoughts during the debrief (and to catch any small errors).

    So I can entirely see his point. It’s reasonable to ask why we put so much time, effort, and money into a system where people with many years of training in mathematics can’t solve problems that rely on middle school math. Does our society really need such a large and unproductive educational infrastructure just so that people like me can learn a lot of math, read a bunch of Latin American literature, have my drawings and designs critiqued by experts, learn a few foreign languages, play with lasers in the physics lab, devolop a strong affection for heap sort, and absolutely fall in love with linear programming?

    (If you want to check your work, I got 26 as the answer to the problem. Since 98!+99!+100! = 98!(1 + 99 + 9900), you can factor 10,000, count how many multiples of 5 are less than 98, and then toss in a few extras for the multiples of 25.)

    Comments
  • Mar 8, 2018

    The Wheel Is a Lie

    Sorry. Not much to report. Plus crazy traffic ate two hours of yesterday because they closed part of the freeway because construction workers punctured a gas line. Plus I had a migraine.

    The person who designed and built the database (as well as all of the pages and tools that it interfaces with) gave me SQL query to extract the important things that I need to know.

    Just by coincidence, the SQL query worked absolutely perfectly on the small test case that I was using to build the proof of concept. I’ve spent much of the past week or so trying to scale things up to deal with the problem in general, including the case where we don’t have any old data to validate against.

    So much time wasted because the query has too many things ANDed into the WHERE clause, so data is missing, but I have no idea how to fix the query.

    Comments
  • Mar 6, 2018

    Pierre Menard, Author of the Quixote

    The five stages of reinventing the wheel

    1. Write the spec for a wheel.
    2. Assign someone to make the wheel.
    3. Make the wheel.
    4. Find the source code for an older wheel.
    5. Use the original wheel.
    Comments
  • Mar 5, 2018

    Another Quilt Done

    While it might be true that yesterday I never showered, got dressed, or left the house, it is also true that I finished another the quilts in progress.

    Quilt

    This one is roughly 84 inches by 84 inches (most of the patches are 6” squares). My artistic inspiration for making this one is that I own way too much fabric and it can be cold in the office. Since I’m keeping this one (in my office), I haven’t yet gotten around to putting in the time to take a reasonable, non-keystoned photograph where the entire quilt is in frame (and square).

    I will leave you with some Monday omens.

    1. Another carryover from Sunday: Walgreens texted me that my prescription refill was ready about an hour before the pharmacy closed, and I was not motivated to drive to Walgreens in my pajamas, so I decided to stop by Walgreens on my way to work this morning. The pharmacy was still closed.

    2. The Starbucks next to the Walgreens was out of my preferred ridiculous drink. A barista claims that every Starbucks west of Bakersfield received incomplete restocking orders this week. Is Amazon secretly running Starbucks as well as Whole Foods?

    3. The Walgreens and the Starbucks are right by the exit to get on the 805. And so I was left thinking about the tragic hero archetype that I learned about in 9th grade English class: Do I get on the 805 north in the morning and then take the 805 to the 163 to the 15? Or do I drive a few miles on surface streets to get on the 163 in Hillcrest or University Heights? My GPS always wants me to take the 805, but I just don’t understand how it can be faster to take the 805 when traffic is almost always motionless.

    4. Our site went down for about 20 minutes today.

    Comments
  • Mar 4, 2018

    Dealing with Users and the Passive Voice

    1. Data were lost.
    2. The server was down.
    3. Posts were deleted.
    4. Accounts violating our terms of service were created from this location.
    5. Restrictions were placed on your account.
    6. The situation is being investigated.

    The passive voice now makes me think of the German news (when the Podcasts app on my phone is behaving reasonably well, I listen to the news in German as I drive to work) because there are so many bits of news about Terroranchlägen, and every single one of them reports the death toll using the construction sind getötet worden. Functionally, that phrase mean “were killed,” but slightly more literally it is “became dead.”

    (German uses the verb werden, “to become,” where English would use a form of “to be” in constructing a sentence in the passive voice. In case you were wondering, werden is a copular verb, so its subject complement is in the nominative case, violating one of the rules that learners-of-German construct when creating their mental model of the grammar. It also, as shown in the example above, takes sein when forming the perfect tense, much to the chagrin of many English speakers. Worse for me, there is only moderate overlap between the sein verbs in German and the être verbs in French.)

    Comments
  • Mar 3, 2018

    Tales from the Sheriff

    Since we are a small company (ONE THAT IS HIRING), everyone needs to do a lot of things. In addition to being the business intelligence lead, and the (homework) BugMaster and triaging all the job applications that come in to our open calls, and helping out with assorted things in the school, I am also in charge of keeping our online community (message boards) from exploding into Lord of the Flies style chaos. (And that really is the right metaphor since most of the participants on our message board are 12-year old boys.)

    1. The MAA still has not released the cut-off scores that determine which students are invited to take the American Invitational Mathematics Exam on Tuesday. That is right, right now it seems like many students will not find out until Monday whether they will be taking a 3-hour exam on Tuesday that they perceive to be important for their future chances of attending Harvard University. Now imagine all these students participating in an internet forum under the cover of pseudonyms.

    2. A user from Kazakhstan was bullying other users on the forum. I cut off that user’s ability to post, temporarily. The user complained. I replied that we just do not allow that sort of behavior on our message boards. He responded with profanity. I extended the ban and also set up restrictions that forbid any new account registrations from just about every IP address in Kazakhstan for the near term.

    3. While updating the notes on a user whose posts get flagged a lot, I wrote, “This user is either as masterful at trolling as Andy Kaufman or else he really is as clueless as he presents himself.”

    4. Relatedly: There are some enthusiastic 11-year-olds who don’t speak English very well who would quite possibly fail a Turing test.

    Comments

subscribe via RSS