ozarque: Linguistics; language as persuasion; part four...

ozarque (

ozarque) wrote,
@ 2007-10-07 09:12:00

Linguistics; language as persuasion; part four...
I'm going to try to make this as clear and as simple as possible; what the chances that I'll succeed in doing that may be, I haven't the vaguest idea, but I intend to try. Because I really do need your help.

Way back in 1989, I came across this paragraph in "Distrust, Rage May Be 'Toxic Core' That Puts 'Type A' Person at Risk," by Chris Raymond, PhD; it was on page 813 of the February 10, 1989 JAMA:

"A study of 255 male physicians, conducted by the Duke team, and a study of 1877 men working at Western Electric have found the same link between hostility and death rates. In the physician study, medical students who scored high had seven times the death rate of low scorers over 25 years follow-up."

[The part about scoring high or low refers to scores on a set of 50 questions on the Minnesota Multiphasic Personality Inventory that are alleged to measure hostility; scorers defined as "high scorers" were those who scored in the upper quartile on that hostility scale.]

I realize that the "physician study" is a very small sample, that it may have contained a dozen wild and wooly variables, that correlation is not causation, and so on -- but bear with me for a moment, please. This is an ongoing body of research. A number of other (and much larger) studies of similar design, with various modifications and improvements, have been done since 1989. I think it's fair to say that it's now no more controversial to say "there's a link of some kind between hostility and heart disease" than it is to say "there's a link of some kind between polluted water and cholera." You'll find an overview of that research, and of its history, in these two articles:

1. "Hostility as a Predictor of Survival in Patients With Coronary Artery Disease," by Stephen H. Boyle, PhD, Redford B. Williams, MD, Daniel B. Mark, MD, Beverly H. Brummett, PhD, Ilene C. Siegler, PhD MPH, Michael J. Helms, BS and John C. Barefoot, PhD --
at http://www.psychosomaticmedicine.org/cgi/content/full/66/5/629

2. "Psychosocial Risk Factors for Cardiovascular Disease: More Than One Culprit at Work," by Redford B. Williams, MD; John C. Barefoot, PhD; and Neil Schneiderman, PhD --
at http://tinyurl.com/2ce3hh

Now, going back to that paragraph from 1989, with all its flaws, here's my question: What are the different ways -- in terms of mathematics and statistics -- to word the statement that medical students who scored high on the hostility scale had seven times the death rate of those who scored low on the hostility scale? Set aside for the moment the flaws in the study and the flaws in the statement, which aren't my concern right now. What I want to know is this: Suppose I just want to tell someone that very crude alleged fact. As in: "Over a 25-year period, you are seven times more likely to die if you score in the top quartile on that hostility scale than if you don't." How many different accepted ways are there for me to word that, and what are they?

There are of course an infinite number of ways to word the statement in the sense of style and vocabulary. But there must be a recognized set of standard ways to word it in the mathematical and statistical senses. I'd like to get a good look at that set. For example, I assume that the set must include a wording like this one: "Over a 25-year period, your chances of dying are [X percent higher] if you score in the top quartile on that hostility scale than if you don't."

The sentence in the paragraph doesn't say anything about cause of death. Maybe it would be easier to construct the set of standard wordings if I did a bit of science fiction here and rewrote the sentence as:

"Over a 25-year period, you are seven times more likely to die of Portmanteau Syndrome if you score in the top quartile on that hostility scale than if you don't."

[So far as I know, there is no such condition as "Portmanteau Syndrome"; I made it up for the occasion.] If that revision would make things simpler, that's fine. I just need to see that set of standard wordings.

Phew. It's very hard to write coherently about something when you don't have any vocabulary for it. If I've asked for the impossible, or my question is incomprehensible, please tell me what I need to do to straighten out the tangle and I'll do my best to follow through.

(Post a new comment)

archangelbeth
2007-10-07 03:07 pm UTC (link)

I'm... Hm. I've been not saying anything 'cause statistics can make my eyes glaze over, sadly. I'm not sure that you can "legitimately" rewrite that sentence to give a cause of death at all -- the premise I'd start with is that hostile people are more stressed than non-hostile people, and that stress creates an overall lowered resistance to other things. So they don't heal as well, don't bounce back as quickly, and/or are more likely to get into a car accident or the like. It's not that hostility causes the death -- as you've said. It's that there's a known link.

Any failings of me to duplicate what you want should be attributed to my "woke up fairly recently, no breakfast yet" brain, not your communication.

Let me see if I can find other ways...

"If you score high on this hostility test, you are 7 times more likely to die in the next 25 years."

"In 25 years, the high-hostility people are 7 times more likely to wind up dead."

"X% number of high-hostility people will die within the next 25 years, while only Y% of the low-hostility people do."

I don't know how to reverse the 7-times into X% in this case. My math, she is rusty at best.

(Reply to this)(Thread)

dpolicar
2007-10-07 04:04 pm UTC (link)

WRT X%, there's two ways to do it.

You can always translate "7x more likely" into "700% as likely".

More interesting would be an objective comparison of % chances. The problem here isn't your math skills; you don't have enough data to do it with. But, since what we're looking for here is structures rather than data... let's suppose that, of the 255 doctors, 122 were high scorers and 122 were low scorers (and one scored in the middle). Let's further suppose that in the 25 years between 1964 and 1989 (or whenever the study was actually done), 2 of the low-group doctors died and 14 of the high-group doctors did.

Given that data, and the simple math 2/122= .0164 and 14/122= .1148, you could say the following:

"Only 1.6% of the low-scorers died in the next 25 years, as compared to 11.5% of the high-scorers."

Or, you could turn it around:
"98.4% of the low-scoring group were alive in 1989, as compared to just 88.5% of the high-scorers."

There's also the time-honored tradition of pretending there's a causal effect, as in: "Lowering your hostility score makes you seven times more likely to live longer." This isn't accepted by ME, but it's accepted by many people who write and read these things.

(Reply to this)(Parent)

Restating your case

thetimesink
2007-10-07 03:08 pm UTC (link)

Not being a statistician, but trying to help, I'll attempt a restatement:

You'd like a language lesson, specific to the Worlds of Mathematics and Statistics, that translates, "Given the specific condition 'Y' (top quartile in 'Study A'), condition 'Z' (death) is 'X' times more likely for a member of that top quartile than for someone not in that quartile." ...and you would like that statement expressed in as many valid ways as is possible within those two disciplines.

Off to read; cardiovascular is an area I'm familiar with; and being male, over fifty, overweight, statin controlled, with two teen-aged males as an uncontrolled local genetic experiment, of more than passing interest...

(Reply to this)(Thread)

	Re: Restating your case thetimesink 2007-10-07 03:13 pm UTC (link)
	...and what Elizabeth said, "Any failings of me to duplicate what you want should be attributed to my "woke up fairly recently, no breakfast yet" brain, not your communication." Coffee, food and increased awareness of the outer world may improve things... (Reply to this)(Parent)

	dargie 2007-10-07 03:31 pm UTC (link)
	"Portmanteau Syndrome" is being crushed under the weight of all the huge suitcases you carry when you travel because foreign countries are so, well, foreign that you need all your stuff. (Reply to this)

ingulf
2007-10-07 03:37 pm UTC (link)

Hmm. Well, from a frequentist perspective, both 'your chances of dying' and ' you are seven times more likely to die' are impermissable, because frequentist probability doesn't apply to individuals. A frequentist would have to say, 'those in the top quartile on the hostility scale are ...' rather than 'you'. A bayesian would be able to say the first two phrases, as long as that's *all* they know. Otherwise they would have to add some caveats about how the other factors they know about are likely to affect your chances.

Supposing that you have just taken this hostilty test, either statistician should tell you how the reliability of the test affects your chances.

I suspect that's not what you wanted to know. What else? Well, a common tactic to help the patient understand risks is to compare it with some risk the patient already understands, such as crossing the road, being hit by a meteorite, smoking, etc.
There are going to be some pitfalls there, but I don't really know what they are.

There are two ways of stating a probability that most people understand: as a percentage, and as odds. You can state both the before and after probabilities in either form, or (as you have done) as a ratio, in either form. In the case of a continuing risk like this one, there is also the question of what timescale to use. The death rate over 25 years will sound more impressive than over one year (unless they are dropping like flies), if stated in absolute terms. As a ratio ('seven times more likely'), they would both be the same, as the time effect should cancel out.

(Reply to this)

(Anonymous)
2007-10-07 04:14 pm UTC (link)

Not quite sure what you're wondering... Different ways of expressing the same specific mathematical statement? Or different statistical ways of describing the same basic concept? Depending on how you slice a data set, you can emphasize one aspect or another.

For example, the following statements might be drawn from one set of data, but they are not necessarily equivalent to each other.

"Those who scored in the top quartile on that hostility scale were seven times more likely to die of Portmanteau Syndrome during the next 25 years than those who scored lower."

"Among those who scored in the top quartile on that hostility scale, Portmanteau Syndrome was the third leading cause of death."

"Eighty percent of those who scored in the top quartile on that hostility scale were diagnosed with Portmanteau Syndrome during the next 25 years, compared with 35% of those in the second quartile, 7% in the third quartile, and none in the lowest."

"Of those who died of Portmanteau Syndrome, 90% had scores in the top quartile on that hostility scale."

"Those who died of Portmanteau Syndrome during the course of the 25-year study had an average score on that hostility scale 60 percentage points higher than those who died of other causes."

(Reply to this)

Do we need to communicate with numbers?

journeyrose
2007-10-07 05:46 pm UTC (link)

Can we leave statistics out of it and say something more universal and at the same time more helpful?

You, NAME HERE, have scored high on the hostility list. Studies have shown that people with higher scores are more likely to die of Portmanteau syndrome, more likely to experience fatal road rage incidents, and have higher blood pressures.

You can help lower these chances by taking a deep breath when you feel angry, re-casting situations so that you don't take them personally, etc.

I think the helpful statements would then be something the listener could DO and take away. So the end result would be in 3 months that the listener would say Well, I was more likely to be in trouble if I continued to be hostile, so I am using relaxation breathing techniques at every stoplight now...

(Reply to this)(Thread)

Re: Do we need to communicate with numbers?

dawnd
2007-10-07 09:17 pm UTC (link)

I think the helpful statements would then be something the listener could DO and take away.

I totally agree here.

I think it's also important to make sure that whatever measures you suggest actually verifiably reduce "hostility," that the patient knows how to DO those measures, remembers when to do them and remembers to do them at those times, and that they actually are doing them *correctly*. Otherwise, you've simply given them another reason to feel like they're a failure, and something new to be hostile about.

This is why ongoing groups can be so important, where people can learn and practice new behaviors, and have encouragement, feedback, and follow-up. One of the problems with doctor-patient communications of this sort (IMO), is how the doctors will often inundate the patient with information, not confirm patient understanding, and send them out the door assuming that they actually know what to do and how to do it, and then never follow up. At least, until the next time the patient is in their office, and they ask "have you been doing what I told you?" and when the patient answers "no," they simply label the patient as "non-compliant," and blame the failure of the treatment on that, rather than their own poor communication and follow-up skills.

(Reply to this)(Parent)

Re: Do we need to communicate with numbers?

undauntra
2007-10-07 11:37 pm UTC (link)

The problem here is that it's not necessarily true that those actions would lower one's chances of dying of Portmanteau Syndrome. Why do you assume that the study shows that the death rate due to Portmanteau Syndrome is caused by the hostility score? Maybe the early stages of Portmanteau Syndrome cause hormonal imbalances which tend to increase hostility. Maybe there's a genetic defect which is a risk factor for both Portmanteau Syndrome and for high hostility.

This is part of why it is important to state the results of studies clearly, and in a way that minimizes misreading. It's very easy to misinterpret statistical results by reading more into them than is actually there.

(Reply to this)(Parent)

Re: Do we need to communicate with numbers?

leora
2007-10-08 12:28 am UTC (link)

Except that the helpful statement is backed up by no research. The problem is the lack of causation showed in a correlative study like this. We can say that people who weren't aware of these results and scored high are at risk, but we don't know why.

So, it's quite possible that there is a factor, say some chemical balance or genetic tendency that causes two symptoms - hostility and heart disease. If that is the case, the hostility may not have any causative effect on the heart disease, it may simply be the messenger. And changing the hostility may be shooting the messenger; taking away the symptom that lets you know the person is at risk, while doing nothing to change the risk. It'd be like giving a baby a drug that makes the baby stop crying, but doing nothing to fix the hungry belly or dirty diaper making the baby cry.

Except that I think decreasing hostility is good in its own right, whether or not it improves your risk for heart disease. So, it wouldn't be necessarily bad. But you can't tell anyone that studies show that decreasing hostility will improve their chances with heart disease even thetiniest possible smidgen unless someone does a study that tests people's incidence of heart disease comparing groups that lower their hostility to groups that don't. And preferably, with people assigned to relaxation techniques or whatnot so you don't have the people self-selecting whether or not to control their hostility, because people willing to do that without prodding may be different from those who aren't willing to.

(Reply to this)(Parent)(Thread)

	Re: Do we need to communicate with numbers? journeyrose 2007-10-08 02:27 pm UTC (link)
	Yes, I was simply assuming that there was cause and effect. Which brings into it--how do we communicate these assumptions? (Reply to this)(Parent)(Thread)

Re: Do we need to communicate with numbers?

leora
2007-10-08 11:14 pm UTC (link)

Every time you assume a correlation means a causation a statistics professor dies - quick, clap your hands to bring one back to life!

It is true that correlation does correlate with causation. And it is true that causation does cause correlation. But getting people to read a causation into a correlation with no proof of causation is one of the biggest ways that people lie with statistics. So, implying a causation you don't know to exist strikes me as very dishonest.

I can give you wonderful lists of things that correlate but have no causation between them (they're both caused by the same factor). For example, reported rapes and ice cream sales positively correlate. They are both more likely to happen when it is warm. It also correlates with frog noise. But reducing frog noise will not reduce rapes. Well, maybe it will, I've never tried it, but I doubt it. Overall, it probably correlates negatively, because I suspect frog noise is more common in lower population densitities, and any sort of crime is more likely to happen if you have more people so you have more criminals.

The field of statistics has a lot to offer; it just is so good at leading people astray. Remember, 3 out of 4 statisticians objected to the small sample size of this study.

(Reply to this)(Parent)

	continued, sorry. journeyrose 2007-10-07 05:47 pm UTC (link)
	If you just say something like if you score high on the hostility factors, you are 7 times more likely to die of Portmanteau Syndrome, people can just say "So what"? And not remember the statistic--or not change their lifestyles to avoid the problem. (Reply to this)(Thread)

Re: continued, sorry.

beckyzoole
2007-10-07 07:43 pm UTC (link)

or not change their lifestyles to avoid the problem

Ah! But the study did not say that changing their lifestyle could avoid the problem.

What is miscommunicated to the media and the public in general is what can be done to change one's own individual risk of death.

I don't know if other studies have shown that reducing one's score on a hostility quiz is associated with lowering one's risk of death. It seems like it should, right? But if the study hasn't been done -- and done several times, with similar results each time -- then we really do NOT know this to be true. All we DO know is that if you are the sort of person who scores high on a hostility quiz in your 20s, you belong to a group that has a mortality rate seven times that of the general population over the next 25 years.

And THAT is what really needs to be communicated better to people, I think.

(Reply to this)(Parent)

What are the real odds?

bemusedoutsider
2007-10-07 08:07 pm UTC (link)

If you just say something like if you score high on the hostility factors, you are 7 times more likely to die of Portmanteau Syndrome, people can just say "So what"? And not remember the statistic--or not change their lifestyles to avoid the problem.

Or they can take it too seriously. It would be more honest to say:
"If you stop X, your chance of getting Y changes from one in 1 million to one in 7 million."

I'd think that most people would say "The hell with it, one in one million is good enough for me."

But then I'd also think that being told that taking a medicine for years may give you three more months of life, would be met with the same reaction; and yet the OP showed a majority of people being willing to take it. (Not to get deeply into this, but I think a survey counting responses that people marked on paper at home and returned at leisure -- might show different results than when a doctor in a hurried appointment is recommending something.)

(Reply to this)(Parent)(Thread)

Re: What are the real odds?

leora
2007-10-08 01:08 am UTC (link)

Of course, we have no evidence it decreases your odds at all.

But I think the one in a million to one in 7 million is a good statstical example. It sounds far less impressive than 6 out of 7 cases of Portmanteu could have been prevented had the people followed this course of action...

6 out of 7, that's a lot. But, of course, those 7 people didn't know they'd become cases and millions of people might have to follow that course of action to prevent those 6 cases, and it sucks for that 7th that went through all that effort for nothing.

(Reply to this)(Parent)

bemusedoutsider
2007-10-07 07:47 pm UTC (link)

I'm going to try to make this as clear and as simple as possible; what the chances that I'll succeed in doing that may be, I haven't the vaguest idea, but I intend to try. Because I really do need your help.

I've been staying out of this partly because it all seems way over my head. But if you want more people in the discussion, examples easier to relate to might be helpful. (Risk factor: getting more replies than you want, and a flamewar.)

There have been a lot of 'studies' about the use of cell phones while driving: whether they cause accidents, correlate with accidents, etc. Cell phone use is something many of us are familar with, so we can see when common sense is being stretched. Looking at some of those issues might at least teach us some of the terms of such 'studies' and how they are reported.

If you'd like to try this, I'll find some references from the last such flame war I was in (I'm very sceptical about the anti-cell side).

(Reply to this)(Thread)

Response to bemusedoutsider....

ozarque
2007-10-08 12:51 pm UTC (link)

Thank you for the suggestion, and for the warnings. [With regard to which I can say only that I can't imagine a condition of getting more replies than I want, but I can imagine what a flamewar would be like, and -- as you already know -- I'd prefer not to be involved in one of those.] However, my preference would be to stay with the hostility research, which has a very large current literature and is directly related to the work I do.

It wouldn't actually matter if I had just made up some entirely hypothetical statistical statement like "People who eat broccoli daily are twice as likely to have cavities in their teeth by age 30 as people who eat carrots daily" and asked for the set of accepted wordings for presenting that alleged "fact." But I failed to get that chunk of meaning across... somehow .... and I suspect that trying again would only create even more confusion. I will resist the temptation.

(Reply to this)(Parent)

ingulf
2007-10-07 09:00 pm UTC (link)

Thinking about it a bit more:

One way to split the problem of statistics from the ethical issues is to use examples. We want to convey the information that hostility is a Risk Factor for some disease. Lets suppose that we have a huge list of risk factors for diseases, being run over, etc. Obviously this would be quite a bit of work to collect, but suppose we have it already. Then instead of going into the math, we can say 'hostility increases your risk of death by the same amount as factor X' (the relative version) or 'the risk of death for those who are hostile is the same as those who have risk factor X' (the absolute version). (I agree with other comments that the relative version leaves out some important information.)

Doing this reduces the problem to one which is easier to get to grips with, namely whether mentioning risk factor X fulfils your criteria of ' most ethical, and most informative, and least coercive'.

Dealing with them easiest first:

Most informative: a comparison with factor X will be informative if a) factor X has the same statistical effect on your life expectancy as hostility and b) our target audience is likely to have a strong and *correct* grasp of factor X.
A good example is probably driving. Most people can probably guess reasonably accurately the probability that some driving mistake has of causing an accident. This is a particularly good one, because you can tailor it to get the risk you want: 'what is the probability of X quarter-seconds inattention on the freeway causing an accident?' for different X.

Least coercive: This is probably audience dependent. For example, you might consider it more coercive to mention dietary risks to a woman than to a man. I guess that the least coercive examples would be risks caused by mistakes that anyone might make, and more coercive those that are considered reckless or shameful.

Most ethical: I can't immediately think of any ethical problems that aren't covered by the first two criteria.

(Reply to this)(Thread)

bemusedoutsider
2007-10-07 11:44 pm UTC (link)

I don't think very many people have a quick and correct grasp of 'S seconds of inattention'. Looking for some really simple comparisons, from which we could work back to health factors....

For adjusting for a comparable risk, it might be better to use a time factor, for example: "One minute spent punching buttons on a cell phone equals X minutes spent punching buttons on a car radio."

Or a distance factor: "Driving 5 miles while smoking a cigarette equals X miles with both hands free for the wheel."

(Reply to this)(Parent)(Thread)

bemusedoutsider
2007-10-08 12:26 am UTC (link)

On second thought: there's a problem with this approach if you're wanting to issue a warning against H (eg hostility). If someone has been doing B long enough to become familiar with it, then they've survived so far -- so they're liable to hear "H is no more dangerous than B."

By now a large number of people do believe/feel that smoking is dangerous, so you might work up somthing like "X years of hostility equals Y years of smoking."

(Reply to this)(Parent)

morgan_dhu
2007-10-07 09:08 pm UTC (link)

It is my job is to write reports that interpret the results of opinion surveys for clients who do not necessarily have any kind of statistical backbround, which means that I am always thinking of how best to give the full story on a difference in percentages and what it means.

In the physician study, medical students who scored high had seven times the death rate of low scorers over 25 years follow-up.

I'd never leave a statement like this that involves only a comparison by itself without some kind of grounding in actual results in a report.

In this case, I'd probably add something to indicate the actual rates found in the study, because there is a big difference between "that is, the death rate of low scorers was equivalent to 1.5 per thousand, while the death rate of high scorers was equivalent to 10.5 per thousand" and "that is, the death rate of low scorers was equivalent to 10 per thousand, while the death rate of high scorers was equivalent to 70 per thousand."

Because the original statement is using a comparison that involves a multiplier, a complete understanding of the impact is dependent on the actual value of the smaller term in the description of this statistical finding.

(Reply to this)(Thread)

	Response to morgan_dhu... ozarque 2007-10-08 12:57 pm UTC (link)
	Thank you; this is helpful. And I am very pleased to learn that the job you are doing exists; that's good news. (Reply to this)(Parent)

carbonelle
2007-10-08 04:35 am UTC (link)

Do you remember the article that you linked to earlier regarding the estrogen-supplement/heart disease controversy? The writer pointed out that (just for example) one correlative factor in the kind of people who do well complying with medical diet surveys, also have personality characteristics which may lead to an orderly, and therefore all-around more healthy lifestyle--including a increased social/economic status?

The same may be true for hostility: that is, the kind of personality which scores highly on anti-social indices such as "hostility" may enjoy (!) similar feed-back loops. They have fewer satisfactory social relationships, which increases their sense of disconnection and rage at the universe at large. Their lack of social support networks puts them at risk for survival from routine physical ailments.

Just as a sample "for example:" When my next door neighbor had an operation, another neighbor made sure her pets were fed and excercised, and I regularly brought over healthy "instant meals." Other friends and family members made sure that she stayed rested and didn't have to exert herself until she was quite recovered.

Do you suppose a "hostile" personality, would enjoy such care?

Since such a possibility exists, to use the language of epidemiology, as: "to say "there's a link of some kind between hostility and heart disease" than it is to say "there's a link of some kind between polluted water and cholera."; is surely... injudicious.

(Reply to this)(Thread)

	leora 2007-10-08 09:46 am UTC (link)
	Departing from statistics and slightly from science for a bit - that's actually a really plausible possible mechanism that would involve a certain amount of causation. And if there is a causation and that mechanism is even partly involved, reducing hostility might actually work to reduce mortality - indirectly, as the person gets better at making and keeping friends. (Reply to this)(Parent)

meepa
2007-10-08 06:33 am UTC (link)

"Over a 25-year period, you are seven times more likely to die of Portmanteau Syndrome if you score in the top quartile on that hostility scale than if you don't."

Given only that information, there's very little I can do to that statement. I need to know different distributional information before I can say anything else.

But one issue is that "you" is involved. It seems much more fair to say "Over a 25-year study, persons who scored in the top quartile of the hostility scale were seven times as likely to die from PS as those who scored in the bottom three quartiles".

Anything involving "you" is immediately suspect as it involves inference to a specific person. The "you" wording immediately implies that if you personally hadn't scored in the top quartile you'd be less likely to die, and thus if you decreased your hostility score you'd be less likely to die. This is of course quite likely to be false, and we need to do more studies before we can conclude anything about it.

(Reply to this)(Thread)

Response to meepa...

ozarque
2007-10-08 01:02 pm UTC (link)

Your statistical point is well taken, and accurate, and I apologize for being unclear. My intention was to use "you" in the sense that the more formal writers in this group would use "one." Generic and indefinite "you," not "you" addressed to some specific individual. Like "You have to make yourself clear if you expect to be understood," with the meaning "One must make oneself clear if one expects to be understood."

(Reply to this)(Parent)

Going back to prayer

journeyrose
2007-10-08 02:26 pm UTC (link)

A while back, you had a post wondering about how to portray or to conduct a study on prayer for a sci-fi work. Zenna Henderson's short story, The Effectives, in Holding Wonder, sets this up rather nicely. There is a new disease which is always fatal unless there is an immediate person-to-person blood transfusion. With the blood transfusion, there is a 25% chance that the person will make it. There are 4 centers that treat the disease. A has a recovery rate of 25%, B has a recovery rate of 50%, C and D have a recovery rate of 100%. The only difference? There is a group of people withdrawn from the land, somewhat akin to the Amish. Donors in this group makes it a point to pray for that person while giving blood. A has no donors from this group, B has half their donors from this group, and C and D have all of their donors from this group.

It's contrived and the math is a little too simple, but the set up is trying to make it a bit more scientific. Zenna has bee out of print for years, but if you want a copy I can send you one I've scored from a used bookstore.

(Reply to this)(Thread)

	Re: Going back to prayer leora 2007-10-08 11:16 pm UTC (link)
	My first thought would be, oh those people are withdrawn, that means they're their own breeding group. They probably had a mutation occur or have some higher density of some genetic difference that makes their blood more useful. That very much wouldn't be a good example for the use of prayer, because the subject pool isn't balanced for the prayer and no prayer group. (Reply to this)(Parent)

	Re: Going back to prayer... response to journeyrose... ozarque 2007-10-09 01:18 pm UTC (link)
	Thanks for posting this. I enjoy Henderson's work, and I thought I'd read all of her published short stories, but "The Effectives" is one I've missed. I'm glad to know about it. As for the math being too simple .... there's no such thing as math that I would find too simple. (Reply to this)(Parent)

	fibermom 2007-10-08 04:33 pm UTC (link)
	I know that you're already familiar with the data on how the phrasing of the statment affects people's response to it. So are you wanting a sort of wardrobe of possible ways to phrase a mathematical statement, so that you can pick the one most likely to result in the behavior change you have in mind? (Reply to this)(Thread)

	Response to fibermom.... ozarque 2007-10-09 01:21 pm UTC (link)
	Almost. I want the standard and accepted set of possible ways to phrase the mathematical statement so that I can look at each one and choose the one that -- if I said it to somebody -- is most likely to result in that person being better informed. (Reply to this)(Parent)

luckylefty
2007-10-09 02:38 pm UTC (link)

Your example re-wording actually changed the meaning.

Even if you assume that the entire reason for the death rate difference is an increase in Portmanteau syndrome, and that Portmanteau syndrome is more likely in high-hostility doctors, the sentence

medical students who scored high had seven times the death rate of low scorers over 25 years follow-up.

has a very different meaning from

Over a 25-year period, you are seven times more likely to die of Portmanteau Syndrome if you score in the top quartile on that hostility scale than if you don't.

For example, let's suppose that Doctors die for two reasons: traffic accidents and Portmanteau syndrome. And assume that hostility increases the risk of death by Portmanteau syndrome, but not by traffic accident (this latter is unlikely, but it's hard to construct an example death cause where it's plausible that hostility has no effect whatsoever. But assume for the sake of the example that hostility gives you lots more heart attacks, but hardly any extra traffic deaths).

Now the numbers might be something like this:

Low Hostility

Traffic: 5 deaths
Portmanteau syndrome: 5 deaths

High Hostility

Traffic: 5 deaths
Portmanteau syndrome: 65 deaths.

Now the high hostility doctors have 7 times as many deaths as the low-hostility doctors. But the reason for this is that Portmanteau syndrome is rendered not 7, but 13 times as likely by high hostility.

As long as there are other, non-hostility-related causes of death, "7 times the death rate" is a different statement than "7 times more likely to succumb to Portmanteau syndrome"

The language of mathematics is great for speaking precisely. The downside of this is that one has to be very careful when rewording, lest one unintentionally change the meaning.

(Reply to this)

Username:		Create an Account Forgot your login? Login w/ OpenID
Password:
	Remember Me
	English • Español • Deutsch • Русский…

About

Help

Get Involved

Legal

Store

LJ Labs