RoR: Forum

Will someone who hasn't voted yet chime in to tell me whether they can see the poll results to the left of this poll discussion thread? I have a suspicion that you don't get to see the results until you first vote. Sounds smart to do things that way.

But is it true (for you)?

Ed

Post 1

Thursday, September 4, 2008 - 7:05am

Ed, I have never seen the poll results prior to voting here on this site.

I posted this poll because I managed to complete my schooling all the way through a bachelor of science degree in mechanical engineering without ever taking a formal academic course in statistics.

We did some statistical analysis via regression in some of the laboratory courses, but that proved a poor foundation for the graduate level course in statistics I completed recently.

Given how broad a swath the study of statistics cuts across such a wide range of disciplines, I have to wonder how much of the educated population even has the first clue about how to understand it.

I find this especially disturbing when vast numbers of people accept certain figures uncritically and then vote based on that disinformation.

Conversely, even educated Objectivists need to accept that variances from a mean happen inevitably and prove hard to control and therefore hard to judge. Public policy can sometimes nudge a mean one way or the other, but variances still happen by the nature of existence. So I would like to see some discussion on this as Objectivists sometimes forget to apply absolutes contextually.

I will refrain from naming instances because I want to give others a chance to do so.

Post 2

Thursday, September 4, 2008 - 7:33am

Ed, the statistics did not show for me when I first arrived. I voted, returned to the page, refreshed it and they appeared.

Post 3

Thursday, September 4, 2008 - 8:16am

Yes, and we learned that self-selecting polls don't mean much.

Post 4

Thursday, September 4, 2008 - 8:28am

Peter, I had to Google "self-selecting polls" to get a gist of what you meant. This article gives some clues. It says, in part:

On closer inspection though, The Times's story was very strange indeed. It told of restaurants, car dealers and shops "facing a bleak time" due to the faltering economy. It revealed that "more than a third of people" are worried about losing their jobs. This gloomy picture emerged, said the paper with undisguised pride, from "a new kind of interactive poll conducted by Times online."

Fair enough, but let's look closer at what that meant. It was not a poll which involved anything like the normal market research study, with weighted samples and so on, but a depended on 2,476 people who decided to take part. In other words, they selected themselves. (See the poll questions here under the oh-so-impartial headline, Are you feeling the pinch?)

Could you please elaborate further?

(Edited by Luke Setzer on 9/04, 8:28am)

Post 5

Thursday, September 4, 2008 - 8:30am

In my civil engineering studies we had a course in theory of measurement which was basically statistics, and I took a night school course. I've long forgotten much of the details but the principles stay with you forever and the subject of probability is most useful in everyday life.

Sam

Post 6

Thursday, September 4, 2008 - 8:30am

I echo Steve's experience.

One of the huge problems in understanding statistics is confounding of data. Basically, the sample data should be as much alike as possible except in the variable(s) you want to study. Here is a Wikipedia page on the topic. If you are going to compare two sets of sample data, pay attention to all the relevant differences between the two data sets, not simply the ones you want to study.

I did not explore this webpage much, but I'm confident you could find some good stuff starting there. I know Milo, he is an Objectivist, and he gave two talks at the 2005 TOC Summer Seminar.

(Edited by Merlin Jetton on 9/04, 8:56am)

Post 7

Thursday, September 4, 2008 - 11:14am

A self-selecting poll in which you announce the questions and then ask people to become part of the sample (i.e. to select themselves) on their own initiative. In a well-conducted poll you pick people at random from some suitably large population, such as subscribers in the phone book, beginning psych students at a university or patients at a clinic. You approach them on your intitiative, with questions in hand, and you publish the percentage of respondents as part of your results.

Self-selecting polls are unreliable because they limit themselves to people with strong enough feelings to take time out for a reply. Thus they risk hijacking, packing or (nowadays) spamming. The article you cite is a case in point. Readers who'll take the time to go online are the ones who've lost their job or think they will soon, who strongly favor or strongly oppose the government in power, etc. The loaded headline Greenslade cites only makes the results less reliable. The answers are probably not the answers of a random cross-section.

These surveys are popular on the web because they are entertaining and, I suspect, because they bump up the publisher's hit count and visit duration and consequently his advertising revenue.

(Edited by Peter Reidy on 9/04, 12:05pm)

Post 8

Thursday, September 4, 2008 - 11:15am

Thanks, Steve.

Ed

Friday, September 5, 2008 - 5:02am

In case anyone cares, I have chosen to audit an online course in undergraduate statistics at my local community college just to reinforce the shaky understanding of key concepts I got in my "drink water from a fire hose" graduate course in the subject.

I just received the complete student package ISBN 9780321577863 from Pearson Education. It looks quite excellent. My community college has an entire set of streaming videos that I found quite helpful as well.

Post 10

Saturday, September 6, 2008 - 10:24am

I've taken an Elementary Statistics class. And it was confusing, but I did learn quite a bit despite my confusion.

Post 11

Tuesday, September 9, 2008 - 10:03am