html xmlns="" xml:lang="en" lang="en"> From the archives: I <em>knew</em> it!

Friday, May 26, 2006

I knew it!

Dependent variable: Will he call? (yes/no)
Independent variables:
He_Said: Did he say he would call? (yes/no)
Predate: Number of pre-date contacts (#)
Cost: Cost of date ($)
Bases: Level of intimacy during the date (bases)
Interest: His estimated interest level during the date (5 point Likert scale)


Labels: ,


Blogger grant said...


11:16 AM  
Blogger TS said...

Do you think you might be over analyzing this just a bit? :)

12:59 PM  
Anonymous badluckblueeyes said...

Or maybe after the date he went home, read your blog, found out you have cats, lived in a semi-nude co-op in Berkeley and send out vibes to married guys.

1:34 PM  
Blogger Aaronlane said...

What's your n?

1:53 PM  
Blogger Megan said...

I was waiting for someone to ask me that. Right now, Aaron's the only man dorky enough for me.

2:00 PM  
Anonymous ed said...

Hmm...the predictive value of the model appears suboptimal. Perhaps if you eliminated some of the variables that would seem to have no predictive value whatsoever (such as He_Said), you might get a better result.

3:31 PM  
Anonymous rogelio said...

girlfriend - you're thinking way too much... unless this was just fer fun. Show the twins off and you'll have to beat them off with a stick.

6:08 PM  
Anonymous Anonymous said...

I like the negative correlation between interest and bases...

-Random guy who found this page via MR.

6:18 PM  
Anonymous Anonymous said...

I've been reading this blog for a couple of weeks. You seem intelligent, cute, athletic, outgoing. I couldn't figure out why you're having so much trouble getting laid. Now I understand.

6:46 PM  
Anonymous meganbythenumbers said...

Cox & Snell aren't the only ones that R Square. You give a whole new meaning to getting a girl's digits.

7:14 PM  
Anonymous hamilton said...

Let it suffice to say that this makes you even hotter [in print] than you already were. How you aren't already beating them off with a stick, I don't know. The sheer power of your intellect, personality, and appearance must be stunning them miles off. Thus, your sample only includes those individuals who are relatively impervious to these attributes. And they're deaf-mutes, too.

PS- Please don't drop variables that lack predictive power; if there's a theoretical reason for having it, keep it. Stepwise regression doesn't become you.;)

9:23 PM  
Anonymous Mitch said...

Some of the literature posits that BASES has a non-linear effect on CALL; it's said that the correlation is positive, but only up to a point. It would be interesting to see BASES get broken out into some dummy variables.

Also, it's impolite to ask someone their N. What matters is the person, not how many trials they've performed. Unless I'm badly mistaken, though, you can read it right off of the classification table.

11:46 PM  
Anonymous Thuddmonkey said...

I suspect this sort of behavior serves as a wonderful filter. The sentiments of the second anonymous poster are probably typical, but only of men you wouldn't want anyway. There seems to be quite a camp of us who are simply more smitten as a result of anything dorky you do. Now, if only I lived anywhere near Sacramento...

Of course, now I'm going to go and admit that I'm not geeky enough, myself; Never having done regression, I had to go look up the notation in order to read your results. :)

12:23 PM  
Blogger Bob V said...

If I were to run this, I'd include how interested *you* were. I know my perception of your level of interest would be important if I were in their shoes.

You have to be careful adding stuff though. You're already estimating 6 parameters with just 26 data points.

Hamilton has a good point about it not being appropriate to drop variables solely because we don't like the results, but it'd sure be tempting to nix "his estimated interest level." Does anyone here have any confidence in Megan's ability to estimate and record that correctly?

12:33 PM  
Blogger Chris Lawrence said...

Interesting model; with a good lit review, you might get an A in my methods class.

Bob V makes a useful point... more generally, significance tests for logistic regression are only valid for large n.

And I agree with "hamilton"; predictors used should be based on theory, not p values.

7:04 PM  
Blogger Aaronlane said...

It's nice to know that my mind is in synch with a pretty young previously semi-nudist. I am actually uplifted by my nerdiness. I revel in it.

But I still want to know what your n was.

10:05 PM  
Blogger Mark said...

For such a small sample size, you want to use likelihood-ratio instead of Wald.

Good job with the high S.E. on bases. Making sure you gather a broad range of hetergenous data is great.

The fact that BASES and INTEREST do not correlate indicates something teribly wrong. I'm guessing it is the unreliability of the highly subjective INTEREST measure.

Kill that one, gather more samples, report back.

10:30 PM  
Blogger Chris Lawrence said...

aaronlane: n (at least for complete cases) is determinable from the classification table. I reckon it as 26.

Maybe INTEREST is coded backwards...

8:09 AM  
Anonymous peefer said...

Hmmm ... 'such a low Wald value for HE_SAID and BASES. That tells me two things:

1) you can probably drop these variables from the model.

2) you tend to date horny liars.

11:03 AM  
Blogger Aaronlane said...

I think He_Said is a misnomer. It isnt shocking that it turns out like it does.

You could just as easily replace it with the She_asks variable. Have you ever heard a guy say "Thanks, I won't call you?"

Maybe not even She_asks, maybe She_leads.

There is only one acceptable answer to that question if you dont want to experience a bad time RIGHT NOW.

The possibility of experiencing a bad time at a later date is discounted in the face of the almost certain probablity of conflict in the next thirty seconds.

8:53 PM  
Blogger Cladeedah said...

All hail Queen of the Nerds.

10:56 PM  
Blogger Robert Prather said...

I agree with hamilton and Chris: if theory dictates a control variable, leave it. However, I'm wondering if there is any formal theory for this study, or if it is simply hard-earned experience.

As with hamilton and others, this regression makes you even more sexy. I've added you to my reader.

12:57 AM  
Anonymous Anonymous said...

Some people roll dice, some people flip coins.

9:41 AM  
Anonymous Michael said...

You are captivating by your intensity, in life and in intellect, and your nerdish tendancies only add to your attractiveness. But I'm gathering this is supposed to be a statistical analysis....

I wonder if the proxy nature of some of your variables is concealing a higher level of Type II error than may be statistically evident, given the number of observations. Perhaps the particular sample you have had the misfortune to draw is unrepresentative of the population. (I think Hamilton put this more cleverly.) This is not to suggest that the male populace is better than we are, but perhaps there are some - hopefully more than a few - who are better than what you have found some of us to be. Nevertheless, judging by the comments you have received in past weeks (an entirely unscientific measure, to be sure, but consider some non-quantitative evidence to cross-check your findings), perhaps you need an entirely new sample set or a new sampling frame. Do you look for them or do they find you? And did they ask you about your thesis topic? ;)

9:35 PM  
Blogger Megan said...

Hey y'all,

I can't believe you think that I gathered this data by myself. How can you have read this blog and think that I have been on twenty-six dates in my life? Nope, other people helped, and I should have thanked Bill long before this for his work on the regression.

11:19 AM  
Blogger bobvis said...

Even a year later, I'm still amazed at how awesome this post was.

5:38 AM  
Blogger Megan said...

A year later, I can't believe I'm writing about exactly the same things. Except I was funnier then.

1:00 PM  

Post a Comment

<< Home