U.S. CitiesCity-Data Forum Index
Go Back   City-Data Forum > General Forums > Great Debates
 [Register]
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
View detailed profile (Advanced) or search
site with Google Custom Search

Search Forums  (Advanced)
Reply Start New Thread
 
Old 05-26-2009, 12:05 PM
 
Location: Portland, Oregon
7,091 posts, read 10,487,344 times
Reputation: 4104

Advertisements

Being a stats nerd I often pick through studies linked to news articles that try and show that A and B are related, and often the correlation the article makes is questionable as best....so why are they trusted without a look? Good studies are sound proof to show a good reason, but it seems like people hoist it up and run with it without even given a glance to source.

I remember a study that was hoisted to show red meat caused cancer, which was front page news where I was going to college (Boulder, lots of Vegans liked it because it made them feel superior). Debates were going back and forth till some one bothered to look at the study, it was a study from Sinclair Wyoming about cancer (big petroleum refinery town). It was questionnaire people who had cancer answered about various lifestyle choices, what they did, where they lived...it showed a great (more then 80% I remember) correlation to living closer to the plant and working in the solvents areas to rates of cancer. Down in the notes, there was a correlation of about 5% of the people who had cancer who checked the box to be eating the most red meat.

Or this recent study Consumerist - When It Comes To Charity, Poor Give Too Much, Get Too Little - Charity about donations by income. I thought this poster summed it up well: "This is an exercise in small numbers, with statistical and journalistic spin. 4.3% of $10,531 is $452.83 whereas 2.1% of $158,888 is $3336.65. You could just as easily state that the wealthy donate to charities at more than 7 times the rate than the poor... but since there are lies, damn lies and then statistics, one has to ask how they defined charity and how complete are the numbers?"

Or my favorite, the "Jaws Study" of 95% of shark attacks happen within 10 miles of shore. Well, it's not like they are going to come to your house and bite you...and most people don't go more then that to have a jaunty swim, and most sharks aren't out there because there's less to eat in the open ocean.
Reply With Quote Quick reply to this message

 
Old 05-26-2009, 12:16 PM
 
Location: Victoria TX
42,668 posts, read 71,538,289 times
Reputation: 35864
I believe, unquestionably, that no matter what your position is, you can find a stat that will support it. For example, if an average doesn't support your point, use a median. And vice versa. i use that technique all the time. Sometimes it really shuts people up.

"Evidence is just a concrete form of wishful thinking." --- Brock Clarke.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 02:18 PM
 
Location: 125 Years Too Late...
10,317 posts, read 9,975,609 times
Reputation: 9083
For anyone who has studied math and/or statistics, the limitation of data analysis and the ease of abuse (a trained statistician selectively 'interpreting' the data) are glaringly clear. I actually did a research project on this topic many years ago. Statistics are great, just be careful about betting the farm on what the data 'shows.' As jtur pointed out, there is a wide variety of averaging algorithms, deviation algorithms, and parametric/nonparametric analysis techniques.

Data analysis/statistics certainly can be and are very useful; but at the same time, they can be and are deceptive when used ‘selectively’ or in an overly zealous fashion. And in many cases, it’s not the data analysis technique that is at fault, it’s simply the presentation. Statisticians know how to get a knee-jerk reaction from the public even with a simple graph--and that is before we even talk about ‘cooking the books’ with the mathematical analysis.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 03:46 PM
 
Location: Brooklyn
40,057 posts, read 29,713,783 times
Reputation: 10450
If you read through enough threads here on C-D, you'll see that plenty of people never question any stats presented to them. (And in addition to accepting stats blindly, when polls are involved, many don't stop to ask how many respondents there were. Polls implying the way Americans feel about a given issue, for instance, that have had a thousand respondents, don't seem very accurate in a nation of 260 million).

For myself, I like to quote Mark Twain: "There are three kinds of lies: lies, damned lies and statistics."
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 03:54 PM
 
1,467 posts, read 1,881,012 times
Reputation: 22487
Quote:
Originally Posted by subsound View Post
Being a stats nerd I often pick through studies linked to news articles that try and show that A and B are related, and often the correlation the article makes is questionable as best.....
Is it just me, or do others find it to be somewhat paradoxical that a stats nerd would ask such a sweeping and subjective question?
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 03:55 PM
 
Location: Abilene, Texas
8,746 posts, read 7,509,242 times
Reputation: 55863
Quote:
Originally Posted by jtur88 View Post
I believe, unquestionably, that no matter what your position is, you can find a stat that will support it. For example, if an average doesn't support your point, use a median. And vice versa. i use that technique all the time. Sometimes it really shuts people up.

"Evidence is just a concrete form of wishful thinking." --- Brock Clarke.
Oh yeah, I did that too one time on a project I worked on. It sounded a lot more impressive when I used the median.

I had a stats professor one time say in class that he could prove there's a correlation between brown shoes and IQ levels..LOL There are many ways to tweek the numbers.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 04:16 PM
 
Location: Sandpoint, Idaho
2,880 posts, read 5,071,331 times
Reputation: 3020
The first course I ever taught as a professor was intro stats. Among my favorite references was this book How to Lie with Statistics - Wikipedia, the free encyclopedia. An oldie but a goodie.

Students' default at age 18 is to believe and to think that all arguments are about truth. Hence, they get very passionate about unprovable debates. And their arguments are really non-arguments, just twisted numbers and rhetoric without any logical or mathematical basis. To them, stats are truth...provided they support their POV.

Even the left brained have fallen into zombie like existence with standardized tests. MCQ exams, and plug-n-chug mindsets. They seem to think that by running a reduced-form regression and getting p<.05 that their answers mean truth. One of the best profs I ever had was a statistician/econometrician who was extremely modest about how he formulated his statistics, how he applied them and how his results were to be interpreted. However, once you enter industry, the quality of statistical work is just garbage, with the focus on finding just the right results to support one's cause/funding. In industry, many of better statisticians would ask, "what has truth have to do with anything?"

Example: In my naive days, I once presented a proposal with a buddy to do a statistical study the effectiveness of a certain innercity education program (about a $10-$15 million program). The presentation went well. The response? "Why do we want to know whether the program is not working?" LOL!! In other words, "we want the stats to show our program is working, dummy!"

Yet, there will always be a market for educated people, despite being lost in a crowd amongst the overschooled, overcredentialed, and unethical. I would rather hire a clear-headed thinker who understands the limits of statistics than a quantjock who would rather not deal with the messy aspects of justifying the use and applications of stats.

S.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 04:26 PM
 
Location: Victoria TX
42,668 posts, read 71,538,289 times
Reputation: 35864
The classic example was then the Wisconsin Works program gathered up a bunch of derelicts and trained them to do things like phone you at supper time and sell you condos. The followed up by doing a survey of "graduates" to see how well they were doing. They released the statistical data that a huge majority were on their feet and self supporting. They conducted the survey by TELEPHONE. Which means the entire survey sample consisted the people for whom it had worked so well, they had homes with landlines installed. They didn't ask the telephoneless "retrainee" who slept on a piece of cardboard under an overpass how he doing, because that would have corrupted the desired effect of doing the survey in the first place.

In a rare moment of journailistic investigative inspiration, the Madison Capital Times pointed this fact out, and was, I think, nominated for a Pulitzer Prize for it.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 05:07 PM
 
Location: Sandpoint, Idaho
2,880 posts, read 5,071,331 times
Reputation: 3020
Quote:
Originally Posted by jtur88 View Post
The classic example was then the Wisconsin Works program gathered up a bunch of derelicts and trained them to do things like phone you at supper time and sell you condos. The followed up by doing a survey of "graduates" to see how well they were doing. They released the statistical data that a huge majority were on their feet and self supporting. They conducted the survey by TELEPHONE. Which means the entire survey sample consisted the people for whom it had worked so well, they had homes with landlines installed. They didn't ask the telephoneless "retrainee" who slept on a piece of cardboard under an overpass how he doing, because that would have corrupted the desired effect of doing the survey in the first place.

In a rare moment of journailistic investigative inspiration, the Madison Capital Times pointed this fact out, and was, I think, nominated for a Pulitzer Prize for it.
LOL! People are desperate for short-cuts. They like being ignorant and want the media to pander to their ignorance, which it does.

American media has become cartoonish. Its ugly. I will always be my father's son, raised on Cronkite, Howard K. Smith, Reasoner and Severeid. Imbeciles, bimbos, and marketing automons are just horrible for the vast majority of American youth who know no alternative.

American universities are so watered down it is beyond belief of most parents.

With each year I taught, things got more frightening. The future we face is so much more daunting than choosing which figurehead to assume the office in Washington.

S.
Reply With Quote Quick reply to this message
 
Old 05-26-2009, 05:47 PM
 
Location: Nebraska
4,179 posts, read 9,115,151 times
Reputation: 9523
I don't believe in statistics. I believe in experience. I've participated in as well as written and evaluated surveys, and always - ALWAYS! - the questions are written to elicit the responses desired. I've seen people try to throw out boxes of responses that didn't fit their purpose. I've seen, even as late as the latest election polls, so much manipulation and outright, purposeful lying even when the numbers are POSTED, that I know that statistics are merely emotionally-manipulated numbers. But so many people will defend them deathlessly, that it is ridiculous to even discuss it with them. People who choose to live their lives in denial of reality are soooo easily taken advantage of, and a total waste of my time - unless I'm targeting them in a survey! Then they practically write my checks for me!
Reply With Quote Quick reply to this message
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.

Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.


Reply
Please update this thread with any new information or opinions. This open thread is still read by thousands of people, so we encourage all additional points of view.

Quick Reply
Message:

Over $104,000 in prizes was already given out to active posters on our forum and additional giveaways are planned!

Go Back   City-Data Forum > General Forums > Great Debates
Follow City-Data.com founder on our Forum or

All times are GMT -6.

2005-2018, Advameg, Inc.

City-Data.com - Archive 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 - Top