[O.C.T.O.B.E.R. IV-8] Attrition (#TBT; Sept. 19, 2007)

I may be in trouble by posting this as a plagiarist, as this is a transcript of my favorite lecture from PSYC-350 in Fall 2007. (Granted, does that mean that since more than 7 years have passed, I won’t get audited?)

I suppose I’ll post it and if I get a Cease and Desist, that will be the case. All right, here we go. I’m going to go put on an orange polo and short khaki shorts… and NOT show the picture because I own neither of those. Ha!

In case you’re wondering, this is from September 19, 2007, and indeed happened right around the time of this post being scheduled to publish.

“All right, attrition. Also known as drop-out, data loss, response refusal, or one of my favorite terms, experimental mortality.

Sounds awful, doesn’t it? Experimental mortality came from the 30’s and the 40’s where most of the research was done on rats, and if you’re a rat, the only way to get out of a study–I mean you can’t no-show, can’t have Mom call, ‘k? Cars don’t get broken down…

… is to DIE! Now there is one more term: experimenter mortality.

[laughter]

And again, you don’t see it that often–that’s when the RESEARCHER dies during the study.

[laughter]

And I’ve only ever had that happen once–it wasn’t to me…

[laughter]

Well, it was to me, in, uh, in a sense–I, I, I went from the graduate student to… the researcher. It was a pretty quick promotion.

And we actually, I, I had it in the manuscript! I talked about how the protocol had to change between the third and the fourth week, because… the first author (still got first author credit, DEAD! He was still the first author! I’m way too nice…)

[pause-chuckles around]

I actually had in there that due to experimenter mortality [laughter in classroom], the interviewer changed between the third and the fourth wave, and they caught it. They caught it. I’ve tried several things over the years and sometimes you can’t get ’em in. I tried three times to refer to the stimuli as having been painted “traditional psychology black.” ‘Cause for some reason if you’re, if you’re playing with shape stimuli, you paint ’em black. Even if they’re not gonna look at ’em. Even if it’s a touch study! BLACK. Has to be black. But they didn’t want me to refer to them as “traditional psychology black.”

[pause]

ATTRITION. ENDANGERS. INITIAL EQUIVALENCE of subject variables. This could be tough–pay attention!

(Why is that down?)

[pause–messing around with the projector]

Ouuhhhhh…

[ding]

[projector turns off and screens retract into the ceiling]

NONONONO! S**T!

[class laughs, and Noah laughs uncontrollably]

[Cal tries to turn the projector back on–beeping is heard, but nothing happens]

Now I have to wait five minutes before the damn thing’ll come back on; I think we’re gonna try and load up this other one…

[laughter and coughing from class]

Some days are better than others. All right.

[coughing from laughing in class]

[writing on whiteboard] We’ve got this pop… [marker is dead]

[laughter as Cal throws the dead marker into the room’s corner]

See? I ordered a couple the other day.

So you’ve got this population. You select… from the population to get a data pool–participant pool. You randomly assign, right? Now we’re hoping, we’re counting on this random assignment to make these two groups equivalent. Before we manipulate the IV, we have initial equivalence of all subject variables, right?

Now, most of the time, one of these… let’s say we got a treatment condition, we got a control condition. The treatment condition requires that these participants engage in a series of practices… before they get tested. The control condition: no practice, just test. Now stop and think: which of these two conditions is more likely to have attrition: the one that takes less time–you go in, you get tested, or the one you’re taken in, tortured for a while, and then tested? Which group is more likely to have attrition?

[Class: “Tortured,” not all at the same time.]

So we’re gonna have more attrition here. [writing] Does that make sense?

Now. We randomly assigned people. So on average, the motivation of the two groups is the same. Then we start manipulating the treatment group.

WHO is likely to attrit, or to bail, from the treatment group? The more motivated or the less motivated?

[Class: “Less motivated.”]

Less motivated. [writing] So if the less motivated people bail out of HERE, and the groups were equivalent in motivation before this started, do we END UP with a difference in motivation between the two groups?

You with me? And since motivation [writing] is a SUBJECT variable, we’re talking about initial equivalence. Does that make sense? The procedure caused the attrition, but the attrition caused a difference in subject variable equivalence between the groups. So attrition is ABOUT. SUBJECT. VARIABLE. CONFOUNDS. Does that make sense?

[pause]

[beep]

[beep; projector comes down]

[beep]

[beep-beep-beep-beep-beep]

[beep; screen comes on]

Oh, there we go. Can you believe that worked? I mean, you push it once, and it doesn’t work, you push it once, and it doesn’t work, [repeatedly pounds on the lectern], and it works? That’s wrong. I’ve just been reinforced for something that I should never be reinforced for!

[pause]

(Is it? OK. Good.) So, attrition works much like “self-assignment” to trash initial equivalence. They both involve a non-random determination of who provides data for what condition of the study. All right? And then, we used my example instead of this one because this was dying.

How to combat attrition: the first thing to do is BEG.

[giggles from the class]

Now, often beg and educate are really the same thing. So educate particabouts… participants about the importance of random assignment to the validity of the study. If there is a differential value of the differential treatments or conditions, offer folks an opportunity to participate in the preferred condition after data collection. That’s BRIBE.

We start with beg, we move to bribe.

Replacement of participants who drop out of the study. And collect data about possible confounding variables for statistical replication. Well, look at this.

[beep; brings another projector down with another screen on]

I did that on purpose. We decided, that in this situation, the less motivated people are gonna come out of here, right? So then we go back to our data pool, we go back to our population, we draw some more, we put ’em in here. Are the people that we put back in ALL gonna be low-motivated?

No–they’re gonna be motivated on average. Which means that this group [i.e., treatment], even with replacement, is gonna be more motivated, ’cause we kept the high-motivated and added back in a mix. Here [i.e., in the control group] we had the mix all along. So replacing attrition doesn’t really get rid of the problem–what you’ve gotta do is go with beg and bribe!

You gotta create a situation where you BEG them to stay the whole time, and you sometimes BRIBE them.

[turns off projector and screen comes up with a whirring noise]

Now, these are not the words you wanna use in public, ‘k? EDUCATE, and ENCOURAGE. ‘K?

But come on, we know what we’re doin’–we’re beggin’ and bribin’.”

======================================================

Today is the eighth day of the fourth round of O.C.T.O.B.E.R. That makes one week and one day.

Memorial Stadium: 16 days.

Thanksgiving Day: 49 days.

Joint Mathematics Meetings: 90 days.

blog signature

Advertisements

Let's have a conversation!

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s