Month: May 2007

Science in Action: Omega-3 (flaxseed vs nothing 2 continued)

Seth Roberts4 Comments

Flaxseed oil increased how fast I did simple arithmetic problems (e.g., 7+5, 9-4, 3*7). To better measure the effects of fats on my brain, I wanted to find out which problems were most sensitive to flaxseed oil. Then I could hope to create a more sensitive test.

Before looking at the data, I assumed that problems that required more processing — more time — would be more sensitive. But this was not what I found.

First, I compared problems with different functions — plus, minus, and times. (E.g., 5+4 is a plus problem.) Sensitivities:

function….F…..n
plus……..15…..1400
minus…….9…..1400
times……16…..1400

The F values are for the flaxseed/nothing comparison. Greater F = more sensitive. The n values are the number of trials. These results more or less agreed with my preconceptions: times problems were slower than the others.

Then I compared problems based on their correct answers. I divided the problems into groups with roughly equal number of trials. Sensitivities:

answer…F……n
0-3 …..12….1000
4-6……15….700
7-9……..6….900
10-19…..2….900
>19……13….500

This was the surprise: Problems with answers 10-19 contributed almost nothing to the overall sensitivity, while problems with answers on both sides contributed much more. Was typing “1″ the problem? No, problems with the single-digit answer “1″ were no worse than problems with similar one-digit answers.

I redid the analysis omitting problems with answers 10-19 and found more sensitivity to the effects of flaxseed oil — a slightly larger F (or t) value, even though the number of trials was 20% less. Here is a graph based on the more sensitive analysis:

new analysis

No important differences from the earlier, less sensitive analysis:

earlier analysis

Based on these results I will get rid of the subtraction problems and the problems with answers from 10-19. I haven’t decided whether I will keep the total number of problems the same; I might increase the number (which is now 100 — 100 arithmetic problems per session).

Andrew Gelman on Blogging (part 3 of 3)

Seth RobertsLeave a comment

What have you learned about blogging?

I’m surprised that there are no other statistics blogs. Chris Genovese used to do one but he stopped–and his blog was more personal than mine, less about statistics. I’ve tried to stay on topic (most of the time), since that’s what I have to offer. I have political opinions, cute stories about my kids, etc., but why should anyone care about this?

I’m also surprised there are so many blogs about economics. Especially since psychology is more interesting (to me). Is this just path-dependence, or maybe actually there are a lot of other blogs out there that I don’t know about. I read some econ blogs partly from clicking my link to Alex’s Marginal Revolution blog which links to others. There are also a lot of tech blogs and computer science blogs but this is less surprising given that it’s on a computer. I suppose there are also lots of blogs about current TV shows and so forth that I don’t really have interest in. As it is, I already read too many blogs. I’m trying to spend more time reading the newspaper. Regarding my own blog, first I was surprised that my students/postdocs/colleagues had so little interest in posting, second I was surprised at how few comments I get on most entries. I don’t always know what will get comments, actually. I’ve had some success using the blog as a sort of out-box where I can park my ideas, but it’s not a panacea. For one thing, it means I spend more time on the computer, which is hard on my hands and maybe degrades my general work productivity.

It’s fun having 1000 readers a day (whatever that means; as a statistical consumer I’m remarkably uncurious about where the numbers come from. One day my postdoc discovered that we had set a switch wrong on the counter, and it turned out we had 1000 rather than 500 per day) although it’s hard for me to think if that’s a lot or a little. Many of my favorite statistical ideas have had struggles with acceptance (for example, there are still a lot of statisticians, even Bayesians, who fit models without even trying to see if they produce replications that look like observed data) so the proselytizer in me wants that large audience.

Finally, I’ve learned that writing can be easier than reading. As is illustrated by the above response that I’ve spewed out.

Parts 1 and 2 of this interview.

R Programs Used to Study Omega-3

Seth Roberts3 Comments

I recently posted that flaxseed oil affected how fast I did simple arithmetic problems (a way to measure brain function). Someone asked to see the R programs I used. Here they are.

> arithmetic.add
function (trials=100, condition = “test2″, procedure = “no mistake correction”,note=””)
{# collect new arithmetic data
#
invisible()
start.time=Sys.time()
condition=new.condition()
start.expt()
trls=sample(rep(c(“+”,”-”,”*”),times=c(34,34,34)),size=trials)#determine trial sequence
t=arithmetic[1,]for (tr in 1:trials){
t2=c(current(),condition=condition,procedure=procedure,trial=tr,arithmetic.new.line(func=trls[tr]),note=note)
t=rbind(t,t2)
}
arithmetic<<-rbind(arithmetic,t[-1,])
save.ws()
arithmetic.plot()
cat(“total time”,round(difftime(Sys.time(),start.time, unit=”mins”)),”minutes\n”)
}

> new.condition
function (conditions.so.far = arithmetic$condition)
{# get new condition name
#
# conditions.so.far vector of conditions so far
#
cat(“most recent condition”,last(conditions.so.far),”\nthis condition”)
condition=scan(nlines=1,what=”character”,quiet=TRUE, sep=”!”)
condition
}

> start.expt
function ()
{#wait for Enter to start data collection
#
cat(“press Enter to start “); scan(quiet=TRUE)
}

> arithmetic.new.line
function (func=”+”)
{# give addition, subtraction, or multiplication problem
#
# func function
#
b=sample(0:9,2)
if(func==”+”) answ=b[1]+b[2]if(func==”-”) {b[1]=b[1]+b[2];answ=b[1]-b[2]}
if(func==”*”) answ=b[1]*b[2]start.time=Sys.time()
cat(b[1],func,b[2],”= “)
a=scan(n=1, what = “integer”, quiet = TRUE)
msec=as.integer(1000*difftime(Sys.time(),start.time,unit=”sec”))
correct=a==answ
if(!correct) print(correct)
list(type=func,first.num=b[1],second.num=b[2], msec=msec, answer=as.integer(a),correct=correct)
}

function (vec, n = 1, drop = 0)
{#return last n elements of vec
#
# vec vector
# n number of elements
# drop elements at end to drop before taking n elements
#
le=length(vec)
okay.le=le-drop
vec=vec[1:okay.le]le=length(vec)
ind=(1+le-n):le
vec[ind]}

If you would like help using these programs, please contact me.

Science in Action: Omega-3 (flaxseed oil vs. nothing 2)

Seth Roberts3 Comments

Here are more results from my comparison of flaxseed oil with nothing. I wrote earlier about the balance results.

Each day I measured how quickly I could do each of 100 simple arithmetic problems (e.g., 2+7, 5-2, 5*6). I learned about this task from Tim Lundeen. Tim did the task with paper and pencil whereas I used R (a statistics-oriented language) on my laptop to do the experiment: For each problem, I typed the answer and then hit return. “FALSE” was printed if I made a mistake. The next problem appeared immediately.

Here are the average solution times:

flaxseed oil vs. nothing

There was a very clear difference, t = 6. It took more than one day for the effect to become clear, in contrast to the balance results, where it took more than two days.

These results confirm what Tim found. Unlike what he did and my balance measurements, no special equipment was used (besides a computer). This may be the easiest way to study the effect.

My balance results have obvious practical value: Old people often seriously hurt themselves due to loss of balance. What about these results? Well, failure to react fast enough causes many car accidents. Flaxseed oil (and probably other sources of omega-3) may make you a better driver.

Directory of my omega-3 posts.

Five Similar Words of Wisdom

Seth RobertsLeave a comment

1. I have mentioned several times what Loren Berlin, a student at the University of North Carolina, told Nicholas Kristof, the New York Times columnist: Stop writing about African failures, start writing about African successes.

2. At the recent New Yorker Conference, Cory Booker, the mayor of Newark, NJ, told this story:

I moved into Newark around 1995. . . In my first month there I saw my first shooting ever. . . . I had my life threatened . . . That same month I met this woman who changed my life. She’s an ornery, tough as nails, just an amazing certifiably insane leader. She was the head of the Brick Tower Tenants Association. . . . I meet this woman . . . The first thing I say to her, in my Yale Law School arrogance, I say to her, “Ma’am, I’m Cory Booker, I live across the street, I’m here to help you.”

She looks at me and she says, “You want to help me, first tell me what you see around you.” . . .

“I see drug dealers.” Which I said in a very respectful tone, in case they overheard me. “I see a crack house.” I describe the neighborhood.

“Well, you could never ever help me.”

“What do you mean?”

“Boy, you need to understand something. The world you see outside of you is a reflection of what you have inside of you. If you only see problems and darkness and despair that’s all there’s ever going to be. But if you see hope and opportunity and even love, then you can be somebody that makes a change.”

3. As I mentioned earlier, in First, Break All the Rules, the authors summarize what they learned from thousands of interviews into one lesson for managers:

Don’t waste time trying to put in [your employees] what was left out. Try to draw out what was left in.

4. At the end of my long self-experimentation paper, I wrote:

[Jane] Jacobs (2000) argued that caste systems and other forms of discrimination retard economic development because they prevent certain jobs from becoming the seeds of new businesses. . . . Belief that something is bad makes it harder to learn what it is good for — including what it could become.

I was referring to the belief of many psychology professors that self-experimentation is bad.

5. In an interview with someone from Buffalo, NY, Jane Jacobs said how development of Buffalo should proceed:

You need to do something — I hate to keep repeating myself — that’s unique to Buffalo, that comes out of Buffalo itself. You don’t want to keep acting like a company town.

In other words, don’t try to make Buffalo more like other cities. Try to make it less like other cities.

Curious, huh?

Thanks to Tobian, who blogs about Ethiopia, for reminding me of the Loren Berlin letter.

Science in Action: Omega-3 (flaxseed oil vs. nothing 1)

Seth Roberts4 Comments

Do our brains need more omega-3? I have blogged about this many times. My first two self-experiments to answer this question compared flaxseed oil (high in omega-3) with olive oil or sesame oil (low in omega-3). My balance was better with flaxseed oil, suggesting the answer is yes. However, another interpretation is (a) flaxseed oil had no effect and (b) sesame oil and olive oil made my balance worse. To test this possibility, I compared flaxseed oil to nothing (no supplement). If flaxseed oil has no effect, the two conditions should be the same. If flaxseed oil improves my balance, my balance should be better during the flaxseed-oil condition.

During the flaxseed-oil condition, I drank 4 tablespoons/day of flaxseed oil — 2 at 10 am, 2 at 10 pm. The balance test was at 8 am the next day. During both conditions, I did not eat fish.

Here are the results:

flaxseed oil vs. nothing
There was a large and very clear difference between the conditions. It took about three days of no flaxseed oil before the difference stopped increasing. On the first day of resumption of the flaxseed oil, my balance was much better than the day before. Comparing the two conditions (omitting the first three days of the nothing condition), t = 7.

These results support the idea that flaxseed oil made my brain work better.

Are injury-causing falls the new scurvy? The large fast improvement in my balance when I resumed flaxseed oil does resemble the large fast improvement when a person with scurvy eats oranges.

I was surprised by the time course of the decrement during the no-flaxseed condition: It looks different than what happened when I drank olive oil. In this experiment, my balance got worse for about 3 days and then stopped getting worse. In the previous experiment, my balance appeared to get worse for at least 9 days. This may due to the high omega-6 content of olive oil — omega-6 (almost identical to omega-3) may displace omega-3. In the absence of omega-6, omega-3 takes longer to deplete.

Andrew Gelman on Blogging (part 2 of 3)

Seth RobertsLeave a comment

To me the most interesting effect of your blog is educational — when I read it I feel like I’m getting a painless lesson in advanced statistics. Any idea if it affects many other readers that way?

It’s nice to hear this, but it’s probably like the difference between watching baseball and playing it. A reader feels he or she is getting an education by reading the blog, but you really learn by doing. On the other hand, you (and many other readers) are active data analysts. So I suspect that you’re really learning from your own data analysis. But the blog could be helpful because you go back and forth–something on the blog can inspire you to try something, which then motivates a question which is answered on the blog, etc.

In any case, I certainly help the people for whom I directly answer questions. Years ago I decided it was less effort to answer people’s questions than to say No. (This was back when strangers would email me after reading Bayesian Data Analysis with questions about nonconverging Gibbs samplers and the like.) Anyway, if I’m answering a question anyway, I might as well do it on the blog.

One thing I’ve tried to avoid is the lazy pattern of answering the easy questions and ducking the hard ones. I notice this on some computer bulletin boards (for example, R-help): There are some people who pounce on any easy question that comes up (often to tell people to Read the Manual). But when you ask a hard question, you get responses from a different sort of person. That’s who I want to be. If it takes too much effort to be helpful in this way, I’d rather not try at all.

Part 1 of this interview.

Andrew Gelman on Blogging (part 1 of 3)

Seth RobertsLeave a comment

Long ago, scholars taught. Then they taught and wrote books. Scholarly journals began. Scholars taught and wrote books and articles. Now a few of them teach, write books and articles, and blog. For example, Andrew Gelman, a professor of statistics at Columbia University, whose blog is here. To learn more about this new form of scholarship, I interviewed Andrew.

What led you to start blogging?

I started the blog in 2004 as a way for the students and postdocs in my research group to communicate with each other–the idea was that we would post items on our recent research and half-baked ideas, and it would be an open forum for us to comment on each others’ ideas, also with the opportunity for outsiders to add thoughts. It also seemed like a good way to publicize our work. I decided to post daily, and I figured that on days that I had nothing to say, I could just post one of my old papers. (As it turned out, I actually have a big backlog of blog entries now.)

Have there been any unexpected effects of blogging?

The blog itself developed differently that I expected. My students and postdocs rarely posted on it (except when I went on vacation and explicitly asked them to do an entry per day) so it became much more of my own personal forum. I’ve somehow developed a fairly equanimous “blog personality” in which I can comment on research by myself and others. Beyond that, I wouldn’t say there have been unexpected effects. The most positive effects have been:

  • Commenters pointing me to software and research of which I’d been unaware;
  • Having to type up my vague ideas has forced the ideas into a less-vague state; it’s also helpful to have to justify my thoughts to skeptical strangers;
  • Publicity for my work; I think that my ideas may be reaching more people now than before.

    • But I anticipated all these effects.

    If those three positive effects went away or became small, would you stop blogging?

    If they all went away, and they weren’t replaced by something else positive, then I suppose I’d stop. I do have a feeling of accomplishment from publishing every weekday for over 3 years (for my own sanity, I generally keep a no-weekend-posts rule), but if I wasn’t getting anything out of it, I’d probably lose motivation and stop.

    Will It Last? (part 2)

    Seth Roberts1 Comment

    I posted earlier about the recent increase in hits to the Shangri-La Diet forums. Here is an updated graph.

    hits vs date

    My shallow data analysis failed to show me the most interesting thing: An increase in the rate of change. To see rates of change, I plotted ratios: hits today/hits one week before. I used one week because there are strong day-of-week effects: Sunday is different from Monday, etc. Here is a graph of these ratios.

    change in hits vs date

    The green line is no change. Points below the green line indicate decreases; above the green line, increases. The red line is from loess. Here is a close-up that shows the red line more clearly.

    change in hits vs date (close-up)

    Since November, the rate of increase has been increasing. 5%/week would be huge; the rate is now 10%/week. This thrills me. It is a sign of what the Chinese call “word to word.”