The value of statistics?

[eve_prime]

In another comm I came across a fun conversation, but I'm not allowed to participate there, and I think it would be interesting to discuss the issue more broadly, so here you go

Comments

[snowspinner]

It seems to me misleading to suggest that "a literary approach" is anywhere near as coherent a formulation as you suggest.

I tend to think any such formulation needs to stop by Stanley Fish's note that some approaches "relieve me of the obligation to be right and demand only that I be interesting."

[sensaes]

I've been arguing for decades that Coleridge was a pomo mofo...

[sensaes]

"...and yet when they're reported in the media they are often presented as if they're a stark and simple truth."

This is why we must learn to despise the media.

Actually, fountaingirl had a good post on this theme fairly recently.

[vlion]

Uh.

Imprecise question yields imprecise answer. I find a flaw with the methodology , not the stats.

[baseballchica03]

Hear, hear! I'm all for a discussion of the relative merits of quantitative vs. qualitative approaches, but the final paragraph takes an overly simplistic idea of what more social scientific PO research really is.

Did I ask five people whom I saw looking at cans of tuna at Safeway?
This is not even remotely how academic surveys are conducted.

Did I randomly call 1000 people on the phone?
"Random" in the meaning of calling 1000 people on the phone and a truly randomized or stratified/representative sample are not the same thing.

Did I ask every single person in the state a thousand times?
This is hardly necessary, given probability and the law of large numbers.

OP, this post is highly biased toward qualitative/"literary" approaches and against quantitative. You seem to be presenting the positives of qualitative analysis and the negatives of quantitative analysis rather than even attempting to address both positive and negative aspects of each.

[fixious]

I'm not sure if your example is good for your purposes -- I think it resists anything but a statistical approach.

[eve_prime]

And yet all social science statistics are snapshots of itty bitty nuggets of "information" that reflect some single dimension of life as lived across many people. I'm saying that people using statistics in this context are not necessarily oblivious to subtleties and are often very aware of the limitations of their findings, and yet those findings may still have useful things to say about the human condition.

[pansette]

That was my reaction, as well.

[venetia]

Oh I don't know, I can easily imagine a c/s paper entitled something like 'The Hunger For Tuna: desire, globalisation and commodification in contemporary Oregon'

[elricmelnibone]

I don't think that's what you meant in your question

Isn't this part of the problem? Can we trust statistical evidence if there is chance that the respondent and the surveyor do not agree on what the question means?

[eve_prime]

It's like trusting any other source, I'd say. If it's a superficial question on a topic where many people's answers would be deep and complex, then probably not very much. If it's a straightforward question which most people would understand as intended but some would not, then that "some would not" issue gets included into the "error term," a catch-all for many aspects of human interaction, including survey-taking, that are imprecise. The error term for asking me just a few times about tuna would be much smaller than the error term for asking me just a few times about blueberries, for example.

People who use surveys professionally in a context where they're trying to really learn and understand the subject they're studying will constantly be striving to grasp the ways in which they and their respondents might misunderstand each other.

[elricmelnibone]

Can you give me an example of a straightforward question?

[pansette]

"May I kick you in the crotch?"

I think that one's fairly straightforward. Pretty much a binary option as far as response.