html xmlns="" xml:lang="en" lang="en"> From the archives: Good luck with that, Part the Second.

Tuesday, October 10, 2006

Good luck with that, Part the Second.

I was talking to that guy from the beginning of the summer; he brought up the possibility of meeting the girl he went back to. That sounded familiar, so I went back to check an old email. Then I made a little table:

Two weeks off-pace? Better make that permanent, buddy.


I am, naturally, ashamed to be basing conclusions on such a small data set. My apologies to all of you. I take heart from Seth Roberts, who is a real professor and everything; some of his research has an N of 1. Also, my friend Teddy got a study with an N of 1 published in Nature; he has had no respect for that journal ever since.


Anonymous Anonymous said...

better make what permanent?


10:43 AM  
Anonymous Anonymous said...

If N=1 is an acceptable data set, consider a story of mine. Back in the mid-1990's I went out on four or five dates with a young woman I had met via a dead-tree personals ad. Things went okay at first but the last date was sort of awkward, and for reasons I'm not sure I understood even at the time I stopped calling her.

Just about a year after our last date, after having seen a couple other women with little "chemistry," I started thinking that maybe I had made a mistake letting this young woman get away, so to speak. Too much time had elapsed for calling her to be anything but hopelessly awkward. I sent her a short note saying in effect that I had liked our prior dating relationship and was sorry that it hadn't continued. If she wanted me to call her again, I said for her to leave her name and number on my voice mail, if not, just do nothing and sic transit gloria mundi.

To my mild surprise, she indeed called the voice mail, I called her back, and we ended up dating for a few months. It didn't work out, but I was happy that I had made the second effort.

Iron Rails & Iron Weights

10:51 AM  
Blogger Bob V said...

Seth Roberts's research is of N=1, but his stuff isn't confirmatory research. Rather, he's just trying stuff out looking for results that must later be confirmed. He explores for things which others must test with a larger sample. Until that happens, it is what it is--intelligent speculation based on a single subject.

Actually, I could argue his research isn't *really* N=1, because he is making several observations on a single subject. If N refers to the number of observations and not the number of subjects, his sample size is pretty big.

11:24 AM  
Blogger Megan said...

You're right. I'll never get into Nature with this research. Unfortunately, I'm not on good enough terms with the guy from two times ago to call and ask him for some data points.

11:33 AM  
Blogger Dubin said...

bbhWhy don't you try to purposely engineer this situation again a few more times? Then you'll have more data. Althought I don't imagine that's GOOD SCIENCE.

12:56 PM  
Blogger Megan said...

My new Craigslist ad!

Hi! Are you smart and funny and nice? Are you also helplessly enthralled to a perfectly nice woman whom you have dated off and on for several years? Call her much? Like every day? Because, you know, you are still friends? Just platonic friends, although maybe you still think of her as beautiful and perfect?

I am the woman for you. Date me once and she'll be back. You can squeeze at least another year out of your drama, and inside, you know you would give your left arm for that. Guaranteed results! Must be willing to fill out questionnaire when finished.

1:15 PM  
Blogger Noel said...

What's wrong with N=1? Nothing I hope, or my research is tanked!

But seriously! What BobV said. You can still do good research with N=1. The psychology literature is full of this stuff, usually case studies of people with particularly rare disorders (for example, idiot savants and the like). Megan, I'm sure some of your dates would qualify. Nature beckons!

1:40 PM  
Blogger Pandax said...

An N=1 might work for psychology but not in core sciences or mathematics.

I deal with n=? every day and cringe when people state findings as dogma based on the opinions of less than 10 people. I'm intrigued, do you remember what the paper topic was that got published in Nature?

2:22 PM  
Blogger Megan said...

I do, but I'm trying to be all stealth about him.

2:29 PM  
Anonymous Anonymous said...

Megan: What will you do this time around when someone asks for a picture?


2:50 PM  
Blogger Megan said...

Justin, you are either incredibly consistent and amazingly literal minded, or you are funny as hell and laughing at me.

I will not be posting that Craigslist ad, because I really do not want to be in the same situation again. The next time I post any personal ad? I dunno.

2:55 PM  
Anonymous Anonymous said...

I'm trying to be funny. But often I'm the only one who ends up laughing.


3:00 PM  
Blogger Dubin said...

But where is D A N ???

4:07 PM  
Blogger Bob V said...

N=1 happens in the core sciences too.

If you're an astrophysicist, you might only have observations from one supernova.

If you're a particle physicist, you might be looking at only one collision out of hundreds or thousands of attempts in a particle accelerator that produced what you were looking for.

And there's only one microwave-background-radiation to look at since we only have one universe that is directly observable.

In general though, you're right that there is less research in the hard sciences with N=1. This isn't because the hard sciences are more strict or intrisically different though. It is because it is frickin hard to get data in the social sciences. In the hard sciences, you can pay a grad student to do something in a lab however many times you please for her to do it.

I can't wait till I get a grad student someday.

4:47 PM  
Blogger Pandax said...

True, unfortunately, there are many events (scientific, social, behavioral, etc.) that are a rare. If a disease or disorder occurs once in a million births or a black hole forms once a millenia, you go with what you can get. Cost and time are a pain aren't they?

In general, one must to be able to validate findings. At least with physics, data often can be supported by established theory. If the experiment is not reproducible, no one will find it credible. I think this is especially true when evaluating behavior among a random sample.

I don't know which would be more stressful, relying on a N=1 or testing something over and over and praying the majority come out the same.

5:54 PM  
Blogger Bob V said...

Relying on N=1 is more stressful.

When you do something over and over again and the results don't come out the same, you just redo the trials where the equipment was acting up.

7:53 PM  
Blogger Ben Wolfson said...

If you're using the guy-1 delta to judge the guy-2 delta, don't you have an N of one too?

10:08 PM  
Anonymous Anonymous said...

M, I'm surprised you've used a table to represent things. given your recent posts I thought you would have preferred a 'pie' chart !

Okay, okay, I can hear you guys groaning from this side of the pond.

6:25 AM  

Post a Comment

<< Home