Another Statistics Blunder???

[Scott24x7]

All,
Just found this opening sentance in a statement from Reuters:

Workers who clocked more than 51 hours at the office each week were 29 percent more likely to have high blood pressure than those who worked 39 hours or less, a new study from California has found.

How can you jump/compare 51 to 39, with only a sinlge percentage???? This just feels wrong.



Best Regards,
Scott

"Everything should be made as simple as possible, and no simpler."

 

[SantaMufasa]

Maniac,

Reuters is saying (in other words): "If you work a lot of overtime, then you are 1/3 more likely to have high blood pressure than those who don't work a full work week."

I don't see a problem with their assertion.

Mufasa
(aka Dave of Sandy, Utah, USA)
[I can provide you with low-cost, remote Database Administration services: see our website and contact me via www.dasages.com]

 

[Scott24x7]

Well, it seems odd to me that you are comparing 51 hours to 39 hours, and then saying they 29% less likely to have a high blood pressure... What if you only work 49 hours???


Best Regards,
Scott

"Everything should be made as simple as possible, and no simpler."

 

[SantaMufasa]

Maniac,

The report is not comparing hours to percentages. My rewording, above, states what they are trying to say. They have not disclosed the statistical findings of a 49-hour work week.

For full, precise disclosure of all the findings, perhaps a graph showing the risk factors of high blood pressure for specific numbers of hours worked, from 0 to 100 hours per week, for example, would tell the complete, and most meaningful story.

Their bottom-line statistical assertion: the more hours you work in a week, the higher your propensity for hypertension.

Mufasa
(aka Dave of Sandy, Utah, USA)
[I can provide you with low-cost, remote Database Administration services: see our website and contact me via www.dasages.com]

 

[Scott24x7]

Yeah, I understood their point... I just thought their presentation lacked a certain.... clarity.


Best Regards,
Scott

"Everything should be made as simple as possible, and no simpler."

 

[Lunatic]

Maniac,

There are three distinct groups here:
39 and fewer hours
40 to 50 hours
51+ hours

That they did not reference the 2nd group when comparing groups isn't important and I don't think that the distinction is unclear. However I do agree that it could be stated more clearly. It probably would have been better to replace "29 percent" with "signficantly".

Workers who clocked more than 51 hours at the office each week were 29 percent significantly more likely to have high blood pressure than those who worked 39 hours or less, a new study from California has found.

Does the public really care that there was no statistical difference between people who:
a) Worked 39 or fewer hours and those that worked 40 to 50;
or
b) Worked 40 to 50 hours and those that worked more than 51?

Would they even understand what that means? ;p

 

[rjoubert]

Don't you know that 74% of all statistics are made up???

 

[Mike Lewis]

Isn't this just another example of the fact that people working in the media simply don't understand statistics. Actually, I'd go further and say many of them don't know how to count.

I see examples from my daily newspaper almost every day:

- Headline: "Housing in City A costs more than in City B"

Fact: The average price of a house in A is higher than the average cost of a house in B.

But that only measures the prices of the houses actually sold. In A, maybe most recent sales have been in up-market areas, while in B perhaps there have been a flood of new lost-cost homes coming on the market. B could still be the more expsensive city.

- Headline: "Drivers are heeding advice not to drink and drive".

Fact: Police forces are issuing fewer on-the-spot fines for drink-driving.

All that tells us is the police's success or otherwise in catching drink-drivers. Maybe there's been a big rise in burglaries, so the police are moving officers from road traffic patrols to anti-burglary patrols.

- Headline: "Motorists are angry because XX Street has the highest rate of parking fines in the country"

Well, some street has to have the highest rate in the country. If the parking attendants declared an amnesty for the street in question, presumably the anger would transfer to motorists in the second-highest street. (In this particular case, the street in question was a particularly long street with a large number of parking bays.)

- Headline. "Prices are falling".

Fact: The rate of inflation is falling.

Prices are still going up, just at a lower rate than before.

I'm sure you could all come up with many more examples.

Mike

__________________________________
Mike Lewis (Edinburgh, Scotland)

My Visual FoxPro site: www.ml-consult.co.uk

 

[AnotherHiggins]

TheManiac,

I don't see a problem with the headline. I think it is obvious that they are comparing those who work more than 10 hours of OT to those who work no OT. As Lunatic pointed out, there is a third implied group for which no statistic is offered.

Perhaps it is worth mentioning for non-US readers that here in the States, a 'normal' work week is defined as 40 hours - typically 5 8-hour days (sometimes 4 10-hour days). For non-salaried employees, working more than 40 hours earns you 'Overtime' pay - typically 1.5X your normal pay-rate.

So I think the average working American will immediately know the importance of 51 hours vs. 39 hours.

For full, precise disclosure of all the findings, perhaps a graph showing the risk factors of high blood pressure for specific numbers of hours worked, from 0 to 100 hours per week, for example, would tell the complete, and most meaningful story.

I imagine that the stress level of folks not working at all would be quite high!

[tt]_____
[blue]-John[/blue][/tt]
[tab][red]The plural of anecdote is not data[/red]

Help us help you. Please read FAQ181-2886 before posting.

 

[gbaughma]

Don't you know that 74% of all statistics are made up???

Yes... and there are only 10 kinds of people in the world.

Those who understand Binary, and those who don't.



Just my 2¢

"In order to start solving a problem, one must first identify its owner." --Me
--Greg

http://parallel.tzo.com

 

[AngelB]

I always thought there were 3 kinds of people in the world:

Those who can count and those who can't


Geraint

The lights are on but nobody's home, my elevator doesn't go to the top. I'm not playing with a full deck, I've lost my marbles. Barenaked Ladies - Crazy

 

[jebenson]

I think the major problem with the Reuters article cited in the OP is the statement: were 29 percent more likely. Well, were 29 percent more likely than what? If I don't know the likelihood of having high blood pressure when one works 39 hours per week, telling me that the likelihood is 29 percent greater at 51 hours per week is just meaningless number-spouting. This particular issue is prevalent in the use of statistics in the media in general. It is done this way because using the actual numbers probably would not provide the "punch" necessary to catch a reader's eye. If the article had stated that people who work 39 hours per week have a 5 in 100 chance of developing high blood pressure, and people who work 51 hours per week have a 6.45 in 100 chance...well, that's just not as "sexy" as "*gasp* OH MY GOD THERE'S A 29% INCREASE!!!!!"

Basically, it's nothing more than fear-mongering.


I used to rock and roll every night and party every day. Then it was every other day. Now I'm lucky if I can find 30 minutes a week in which to get funky. - Homer Simpson

Arrrr, mateys! Ye needs ta be preparin' yerselves fer Talk Like a Pirate Day! Ye has a choice: talk like a pira

 

[traingamer]

The original statement was crystal clear and precise.
Lunatic's assertion/ assumption that

Does the public really care that there was no statistical difference between people who:
a) Worked 39 or fewer hours and those that worked 40 to 50;
or
b) Worked 40 to 50 hours and those that worked more than 51?

can not be logically derived from that original statement.

Maybe we need a study to determine if those working 51 or more hours per week and also post on TekTips have lower blood pressure than those who do not.

Greg
"Personally, I am always ready to learn, although I do not always like being taught." - Winston Churchill

 

[gbaughma]

Well, were 29 percent more likely than what?

I had a candy bar this morning, and right on the front of it it said "Still 30% less fat!" and I thought to myself, "30% less than what? Pure fat? I'll bet it's not 30% less fat than something WITHOUT fat."

I mean, what a crock... and how could it *STILL* be 30% less fat? 30% less fat than the last time they checked? Wouldn't that be like 45% less fat now? Eventually, would it be 100% less fat, because each time they check, it's less and less fat?

Then, I'm looking at a Hershey's carb bar (that I'm eating right now)... So, I'm reading the back of it (never a good thing to do... you don't *really* want to know), and even though it's called a bar for "low carb lifestyles", the back says it has 18 grams of carbs. And <1 g of sugars. By the same token, it's 11 grams of fat, and 7 grams of saturated fat.

So.... I keep my carb level down, but I clog my arteries with fats.....

You know, talk to an old farmer some time. You know the type... 92 years old, and *still* driving the tractor... skinny as a rail but muscular. He'll tell you. He's the guy who ate a half-dozen eggs every morning with a half pound of bacon, as well as a good strong cup of coffee in the morning. He never worried about carbs, or fat, or (bob forbid!) cholesterol in those eggs he was eating... because after breakfast he went and spent 14 hours on the tractor. And he'll outlive all of us.

Diets are simple..... Calories Burned > Calories In = Healthy.

Calories In > Calories Burned = Unhealthy.

(How on earth did I get on this rant?????)

--Greg (Now, with 30% FEWER BRAIN CELLS!)



Just my 2¢

"In order to start solving a problem, one must first identify its owner." --Me
--Greg

http://parallel.tzo.com

 

[Lunatic]

Jebenson

The actual amount of the increased risk is not necessary. You don't have to know the actual numbers. If the group working <40 hours has a risk of X (or for purposes here 1X) then those working >50 hours hour has a risk of 1.29X. Regardless of what the actual values are, there is an increase of over 1/4 in the occurance of whatever is being measured (high blood pressure in this case). The point is that if you work longer hours, you have more risk of high blood pressure. Its comparing the risk between those 2 groups, not comparing the actual risk posed to each group.


traingamer

True, I realized after I posted it that there was no evidence one way or the other. And as of yet I still can't find an edit feature

That was just an instinctive reaction that if you don't show a statistically signficant difference then you can't really say anything about it. I didn't factor in that it may just not be 'sexy' (as jebenson put it) enough to the media.

 

[Golom]

In addition to the missing stats on the 40-50 group there are also lots of other "missings" in this.

- Do the two groups have the same age distributions?
- The same diet and exercise frequency?
- The same gender distribution?

Another favorite trick of the statistically illiterate is abusive use of Relative Risk.

Suppose that each sample has 100 people in it and the findings are
[tt]
Group # with elevated BP
<=39 5
>=59 6
[/tt]
Then the relative risk is
(0.06) / (0.05) = 1.2

and the headline says

"People working more than 59 hours per week
have a 20% higher incidence of high blood pressure."

Proper statistical handling says that a relative risk less than 2.0 (or greater than 0.5 if we're looking at a decrease) should be ignored because it could be caused by random sampling errors as easily as a real trend in the data.
Is 6 out of 100 really much different than 5 out of 100?

 

[jebenson]

Lunatic

Yes, it is comparing the risk between the two groups. But unless you know the starting point, the degree of difference between the two groups is misleading, at best.

Let's say I give you a "statistic" that "If you do x, you will be 3 times more likely to experience y than someone who does not do x." Well, what are the chances of experiencing y if one does not do x? Is it 1 in 1 million? In this case, "3 times more likely" is 3 in 1 million. Is the likelihood 1 in 10, and if I do x is the likelihood now 3 in 10 - again, 3 times more likely? In both cases the chances have tripled, but a 3 in 10 chance is much more significant than a 3 in 1 million chance. Telling a simple percentage increase, without telling what the original numer was, is meaningless.




I used to rock and roll every night and party every day. Then it was every other day. Now I'm lucky if I can find 30 minutes a week in which to get funky. - Homer Simpson

Arrrr, mateys! Ye needs ta be preparin' yerselves fer Talk Like a Pirate Day! Ye has a choice: talk like a pira

 

[AnotherHiggins]

Just so everyone can see what we're talking about, here is the article.

[tt]_____
[blue]-John[/blue][/tt]
[tab][red]The plural of anecdote is not data[/red]

Help us help you. Please read FAQ181-2886 before posting.

 

[Lunatic]

Golom

In addition to the missing stats on the 40-50 group there are also lots of other "missings" in this.

- Do the two groups have the same age distributions?
- The same diet and exercise frequency?
- The same gender distribution?

If it was a valid study it would hold hold everything else steady. We don't need to know age or gender or diet or exercise because they are essentially irrelvent. The model would hold all those factors steady and instead let you focus on the factor you are looking at, hours worked.

You've got a ways to go to show that there should be some mention of those in the article. (However if none of those were held constant then there would be a problem, but without access to the core paper I don't see how we would know).

***

Relative risk seems fishy to me. It seems like it is almost 100% dependent on the size of the intial risk (which is more of Jebenson's point). And statistically speaking its flat wrong if someone provides you with the proper caveats.

If someone gives you the difference, the confidence interval and the standard error, then that tells you if there is any sampling error.

If I tell you that with a 95% degree of certainty that +/- 3 percent, workers who work more than 50 hours a week have a 29% higher risk of HBP than those who work less than 40 hours a week; then any variation in the sample is already accounted for and relative risk, as you have explained it, would mis-represent the results by ignoring 'minor' variations. Any risk with the proper statistical caveats on it already account for everything you've explained relative risk is supposed to account for.

***

Jebenson

That would be true if it were something we knew nothing about. However given that we know HBP is a definate factor in the world we live in, it does not need to be restated.

That being said, much like Maniac's original point, the article would be a much stronger article if the base risk was identified. The article being stronger is vastly different from the article not being useful.

 

[Golom]

Lunatic

If it was a valid study ...

I agree. Point is, all the news report says is that "... the researchers looked at a representative sample ...".

That could just mean that the group was sufficiently large to be taken as a statistically significant piece of the whole population.

We do not know if that means a sample whose membership was constant with respect to other factors or did they just ask

- How old are you?
- How many hours do you work per week?
- What's your Blood Pressure?

For all we know (and possibly for all the researchers know), the "less than 39 group", in addition to having lower blood pressure are also younger (or older); eat better (or worse); are predominantly female (or male); etc.

Relative risk seems fishy ...

Indeed it does ... but it's a standard statistical measure. It is particularly subject to abuse by those who don't understand it's limitations and who therefore generate headlines like the (admittedly contrived) one I presented.