Help - Need some deceptive statistically data

[Janice in NJ]

Help! dd needs help finding an example for her stats class - can't be advertising. ;)

 

An example from a current newspaper or magazine in which data has been represented in a deceptive manner.

 

(Please include a link.)

 

She poked around for a while and couldn't find anything. She can easily write about the data to do the assignment. She just can't FIND something to write about.

 

Thanks!

Janice

[regentrude]

Did you read the newspaper today? The most recent misinterpretation of data should have made front page news:

 

"47% of US Americans pay no taxes."

 

Misinterpretation of data here:

http://politicalticker.blogs.cnn.com/2012/09/17/controversial-private-fund-raiser-video-shows-candid-romney/?hpt=hp_t1

"There are 47 percent ...who are dependent on government, who believe that, that they are victims, who believe that government has the responsibility to care for them. Who believe that they are entitled to health care, to food, to housing."

 

Actual data here:

http://money.cnn.com/galleries/2012/pf/taxes/1203/gallery.election.moneymag/index.html?hpt=hp_t1

 

Just because a person pays no federal income tax, it does not mean the person

a) pays no taxes at all

b) relies on government assistance

c) voted for Obama in the last election.

 

Hope this does not violate the rule against political posts. But it was the most prominent example I could find that fits the criteria.

Edited September 18, 2012 by regentrude

[regentrude]

Another example that illustrates the confusion between causality and correlation:

http://thechart.blogs.cnn.com/2012/09/17/teens-who-sext-more-likely-to-be-sexually-active/?hpt=hp_bn12

 

"Researchers are trying to better understand if young people are at greater risk for HIV or other sexually transmitted diseases because they are sending sexually explicit photos or text messages via cell phones."

[idnib]

Is she looking for stories in which the subject of the story has misrepresented data (see Romney article above?) Or is she looking for an article in which the newspaper itself has misrepresented the data in original analysis and reporting? (Someone wrote an article after reading some statistic and made a bunch of assumptions.)

[regentrude]

or take any study involving medical issues and human subjects:

http://health.webpeller.com/2009/04/meds-help-kids-with-adhd-in-classroom.html

 

"Of 594 children whose parents reported an ADHD diagnosis, those who took medication scored 2.9 points higher on standardized math tests and 5.4 points higher on reading tests than children with ADHD who were not taking medication."

 

Questions to ask: was it a double blind study? (probably not)

is 2.9 points (out of how many points?) statistically significant?

how many participants did take medication? What is the standard deviation?

 

 

She can probably take almost any psychological study and find that the sample sizes are too small, that it has not been corrected for all factors (family dynamics, socioeconomics) and that the effect can not be causally traced to the variable that was studied.

 

Here is the Pelsser study on ADHD and diet: http://www.thelancet.com/journals/lancet/article/PIIS0140-6736%2810%2962227-1/abstract

"Between Nov 4, 2008, and Sept 29, 2009, 100 children were enrolled and randomly assigned to the control group (n=50) or the diet group (n=50)."

A group of 50 subjects is too small to draw statistically significant conclusions.

Edited September 18, 2012 by regentrude

[idnib]

Or check any sort of story designed to scare people into reading the article. (See: "Eggs are good! Eggs are bad! Eggs are good! Eggs are bad!") Also check things regarding safety, crime stats, government spending, etc.

[Sebastian (a lady)]

There is a series of pro con essays on health topics in today's Wall Street Journal. There was a pair on overuse of ADHD meds that had numbers being tossed around like Frisbees. (I thought the one stating that ADHD wasn't over diagnosed was especially badly written).

[lewelma]

Disclaimer: I am NOT talking politics, I am talking statistics

 

Just because a person pays no federal income tax, it does not mean the person

a) pays no taxes at all

b) relies on government assistance

c) voted for Obama in the last election..

 

I just said these exact 3 points to my husband 20 minutes ago. Learning how to understand statistical data is critical in our society because of all the errors in reporting.

 

Ruth in NZ

[stripe]

Just because a person pays no federal income tax, it does not mean the person

a) pays no taxes at all

b) relies on government assistance

c) voted for Obama in the last election.

 

Or d) is impoverished. Apparently there were 4000 people who made over $1 million last year who paid no income tax.

[dirty ethel rackham]

 

Here is something that has been making the rounds on my crunchy friends' facebook pages. I can't find the source of this graphic. After further research, it was found that the percentages were based upon calories, not upon volume. Take kale for example. You would have to eat 21 cups of kale to get the same amount of protein in 1 cup of cooked chicken.

[momma2three]

Please note that I'm not posting this to be political: it's just a graphic I saw on Facebook several times and googled for it and the top result which had it is a political site.

 

http://mediamatters.org/blog/2012/07/31/dishonest-fox-chart-bush-tax-cut-edition/189046

[Candid]

This is old but I found it interesting when I first read it because I suspect that the first set of researchers didn't mean to present only partial findings, but the clever second set were able to see the other part of the story the numbers were telling.

 

The most thorough synopsis I could find is here: http://www.fnsa.org/apaw/ch13.html

Unfortunately the original story was published in 1981 and the research organization that publishes the journal only has issues back to 1995 online so you can't read the original article.

 

Here's the quote from the page I linked:

 

Sociologist David Greenberg did a study comparing the membership characteristics of the nation's two largest organizations supporting and opposing abortion. The National Abortion Rights Action League (NARAL) and the National Right to Life Committee (NRL) each had about 450 of their members surveyed. Most of what he found wasn't surprising, but he did ask the female members whether they had ever had an abortion, and those results showed that 32 percent of NARAL's women members had, while only three percent of NRL's did.[2] Greenberg then concluded that women who had abortions were more likely to favor having it available. What he didn't look at, though, were the absolute numbers, rather than the percentages.

 

According to the Encyclopedia of Associations, NARAL had a total membership of 156,000, while NRL had 12 million. Granberg found 78 percent of NARAL's membership was female, while 63 percent of NRL's was. So 32 percent of 78 percent of 156,000 gives 39,000 such members, while three percent of 63 percent of 12 million yields 245,000 women who have had abortions. In other words, of those women who have had abortions who then proceed to become active on the abortion issue, six times as many joined the opposition.

 

Obviously, this comparison is only of two organizations but it is interesting how the raw numbers showed another story beyond what the percentages did because of the relative size difference in the two groups membership numbers.

[Sebastian (a lady)]

This is old but I found it interesting when I first read it because I suspect that the first set of researchers didn't mean to present only partial findings, but the clever second set were able to see the other part of the story the numbers were telling.

 

The most thorough synopsis I could find is here: http://www.fnsa.org/apaw/ch13.html

Unfortunately the original story was published in 1981 and the research organization that publishes the journal only has issues back to 1995 online so you can't read the original article.

 

Here's the quote from the page I linked:

 

 

 

Obviously, this comparison is only of two organizations but it is interesting how the raw numbers showed another story beyond what the percentages did because of the relative size difference in the two groups membership numbers.

 

Ah, lies, d*mned lies and statistics.

 

One of my favorites is articles that cite gun fatalities for "children" but that count 18 and 19 year olds as children because they are "teens". (You wouldn't expect to see 18 - 19 year old voters described as child voters. Although you might see young twenty something voting habits described in an article about youth voters, making not distinction between new 18 year old high schoolers and 23 year olds who've been in the military for several years.)

[Sebastian (a lady)]

Please note that I'm not posting this to be political: it's just a graphic I saw on Facebook several times and googled for it and the top result which had it is a political site.

 

http://mediamatters.org/blog/2012/07/31/dishonest-fox-chart-bush-tax-cut-edition/189046

 

Playing with scale is common across many sources. It is a good one to look out for.

 

Another one is when something is described as a cut or an increase, but is really a change in growth or a change in projection. So if there was a five year budget that penciled in spending on cheese filled blinis that increased 5% each year and a new budget suggests that the spending only go up 3% next year, it is often reported as a 2% cut; even though the spending on cheese-filled blinis is still increasing. (If my cable company told me they were cutting my prices by making a smaller increase than projected, I would still think my bill had gone up.)

 

There is a chapter I appreciated in the book The War against Boys that discussed a survey that had an uneven distribution of positive and negative answers. So you might have "Very Happy", "Happy", "Somewhat Happy", "Neutral", and "Unhappy" so there were 3 answers that would be counted as happy and only one as unhappy. I think there were also instances where neutral was counted as "not negative" and others where it was "not positive" depending on how the report wanted to spin the results.

[Candid]

Ah, lies, d*mned lies and statistics.

 

I've always dislike that quote because it seems to show ignorance of the underlying math and thinking. If anything it wants statistics to be too rigid.

 

One of my favorites is articles that cite gun fatalities for "children" but that count 18 and 19 year olds as children because they are "teens". (You wouldn't expect to see 18 - 19 year old voters described as child voters. Although you might see young twenty something voting habits described in an article about youth voters, making not distinction between new 18 year old high schoolers and 23 year olds who've been in the military for several years.)

 

Well yes, my youngest is working through Traditional Logic which starts with the idea of defining your terms. Certainly this is a good example.

[Sebastian (a lady)]

I've always dislike that quote because it seems to show ignorance of the underlying math and thinking. If anything it wants statistics to be too rigid.

 

 

 

I've always thought that it said more about the person using the statistics than about the numbers themselves. More in the sense of garbage in, garbage out.

 

I love answering polls. But I often find myself reflecting in the middle that the questions are so poorly written that the results could be taken to mean two opposite things. (What does "Do you think the country is on the right track or the wrong track" really convey? You could get the same wrong track answer from both a Tea Party and an Occupy Wall Street respondent.)

 

I really need to dust off my copy of Boorstin's The Image and reread it. It's my favorite political season book ever.

[Jen the RD]

While looking on the internet might be "convenient", I'd get her to a good library where she can look at several days or weeks of national newspapers like the Washington Post, New York Times, Chicago Tribune, etc. If it's a college library where she can access some of the peer-reviewed journals that are supposedly being reported on, she can actually examine the original research and see how the media misinterpreted the data. For medical info, you'll often see these journals quoted: Journal of the American Medical Association, Lancet, European Journal of Medicine, American Journal of Clinical Nutrition, Academy of Nutrition and Dietetics.

 

Hope this helps!

Jennifer

[Paige]

Most surveys out about the election/issues/opinion polls are somewhat misleading these days. You can look into who was polled (is the sample disproportionately full of people who favor one side or the other), the margin of error and confidence interval (survey shows Senator X in the lead- 52-48%, margin of error +/- 5%, with 90% confidence interval), crime stats in newspapers can have really bad graphics with time periods tightening to show an effect when if you had plotted out over a longer period, the trend becomes unremarkable.

[Janice in NJ]

You folks are SO great. She is all set now - assignment done.

 

THANK YOU!:001_smile:

Janice

 

 

Help! dd needs help finding an example for her stats class - can't be advertising. ;)

 

An example from a current newspaper or magazine in which data has been represented in a deceptive manner.

 

(Please include a link.)

 

She poked around for a while and couldn't find anything. She can easily write about the data to do the assignment. She just can't FIND something to write about.

 

Thanks!

Janice

[Candid]

I've always thought that it said more about the person using the statistics than about the numbers themselves. More in the sense of garbage in, garbage out.

 

I love answering polls. But I often find myself reflecting in the middle that the questions are so poorly written that the results could be taken to mean two opposite things. (What does "Do you think the country is on the right track or the wrong track" really convey? You could get the same wrong track answer from both a Tea Party and an Occupy Wall Street respondent.)

 

I really need to dust off my copy of Boorstin's The Image and reread it. It's my favorite political season book ever.

 

Yes, but you said it about the study I suggest which I don't think is garbage in and out at all. Rather it shows how difficult social science research can be.

 

I'm not sure I will be able to explain this or not, but what follows is my line of thinking. If it makes no sense, then let me know where I went wrong and I'll try to restate.

 

I studied statistics and research design in a social science department and we often talked about these matters. How sometimes the math itself could come up with answers that ran contrary to the common wisdom because people are inherently complex. My dad was a microbiologist who I quizzed while studying stat and learned he used the most basic of statistics (t tests) to vet his experiments, while over in my department we were just getting computer technology to help run statistical analysis that could never have been done before due to the amount of time to hand calculate results without a computer. Dad's tests had one variable that he studied while in human data sets we had an almost infinite number few of which could be ethically controlled in any meaningful way so we needed more complex math to reflect that complexity. Much like Flatland we need an infinite number of dimensions to reflect reality.

 

I suspect that neither set or folks using the abortion/political activism data set manipulated the data at all instead what the data set reveals is the complexity of human behavior. AND in the case of the first research I suspect it never occurred to him to look at the relative size of either organization and see how his numbers might be impacted by that. If it had, he may have looked to see if there was another larger organization that he should be looking at. Instead much like how a mathematical proof can wait centuries for someone to solve it but once solve it is being taught in high school soon thereafter as totally obvious, so the second set of researchers had that additional set of facts to add to the mix and were able to reframe the data set.

 

I guess you could assume that he knew that the two groups were not the same size and chose them for what he wished to show, but he got caught, but I don't think so. Perhaps I am naive in this belief.