Bill off net

CSGnet Archive CSG2007
1

[From Bill Powers (2007.08.02.0834 MDT)]

Hello, all. Interesting discussions, but I can’t really join right now
because my internet connection at home is down. I’m typing this at
Mcdonald’s.

Brief comment. I keep forgetting this, but my initial question on this
statistics business was very simple: given a certain correlation, what is
the probability that a statement about an individual that is based
on group statistics will be incorrect? The answer to that question is not
an argument about whether we should do research or apply what is known:
it is a number, or a table of numbers. So far only Rick has done any work
toward obtaining that number or table. All the rest consists of people
telling us what they have concluded on the basis of what they have
assumed to be true.

The second comment is similar: can anyone lay out the theoretical basis
for libertarianism, representative democracy, rule of law, dog-eat-dog,
or whatever system concept likes behind the political discussions? We’re
in a theory-based seminar here, so it seems appropriate to ask what
theoretical bases there are for the various points of view.

My replacement DSL model is supposed to be delivered tomorrow afternoon.
I may come back to McDonald’s tomorrow just to check up on
things.

Best,

Bill P.

P.S. Yes, Gary, I have had all the same thoughts about being on the
Mississippi and passing under bridges in Minneapolis.

···

–

No virus found in this outgoing message.

Checked by AVG Free Edition. Version: 7.5.476 / Virus Database:
269.10.23/924 - Release Date: 7/28/2007 3:50 PM

No virus found in this incoming message.

Checked by AVG Free Edition.

Version: 7.5.476 / Virus Database: 269.10.23/924 - Release Date:
7/28/2007 3:50 PM

2

[From Jeff Vancouver (2007.08.02.1545 EST)]

[From Bill Powers (2007.08.02.0834 MDT)]
Brief comment. I keep forgetting this, but my initial question on this
statistics business was very simple: given a certain correlation, what is the
probability that a statement about an individual that is based on group
statistics will be incorrect? The answer to that question is not an argument
about whether we should do research or apply what is known: it is a number, or
a table of numbers. So far only Rick has done any work toward obtaining that
number or table.

This is not a simple question. If the correlation is perfect (1
or -1) then the answer is everyone. Anything less and the question becomes
meaningless. Indeed, a zero correlation could arise when half the sample have a
positive effect and half a negative effect of equal magnitude, or no effect for
anyone (or anything in between).

The second comment is similar: can anyone lay out the
theoretical basis for libertarianism, representative democracy, rule of law,
dog-eat-dog, or whatever system concept likes behind the political discussions?
We’re in a theory-based seminar here, so it seems appropriate to ask what
theoretical bases there are for the various points of view.

Steven Pinder has a good discussion of these issues in his Blank
Slate book. In short, the answer seems twofold: first, individuals have
assumptions/theories that support their belief systems (this is mostly what
Pinder talks about), and second, that individuals do not seem overly concerned with
or capable of discerning internal consistency within their belief system. The
big issue in the second statement is what does “overly” mean. It
appears there is some concern. It also appears there are individual differences
in concern and epistemology (i.e., critical analysis of information sought and
provided). An interesting PCT question is whether we desire consistency and how
we know (perceive) our degree of consistency. Social psychologists seem to have
accepted that we desire it (Leon Feshinger was a big promoter of the concept). However,
it seems highly likely that our input function for creating the perception struggles
or that the gain in the system is low for some and/or some of the time. Indeed,
I cannot imagine doing anything else if all my resources had to be put to making
sure my belief system was consistent (true it did occupy a great deal of my
time when I was in high school – though girls still took precedent even
then). I just don’t think I could ever reduce that discrepancy
completely, though I do bump up gain or reference level from time to time on
that system (I think).

Jeff V.

3

[Martin Taylor 2007.08.02.16.38]

[From Bill Powers (2007.08.02.0834 MDT)]

Hello, all. Interesting discussions, but I can't really join right now because my internet connection at home is down. I'm typing this at Mcdonald's.

Finally, I've heard of a legitimate reason for going into a Mcdonald's!

Brief comment. I keep forgetting this, but my initial question on this statistics business was very simple: given a certain correlation, what is the probability that a statement about an individual that is based on group statistics will be incorrect? The answer to that question is not an argument about whether we should do research or apply what is known: it is a number, or a table of numbers. So far only Rick has done any work toward obtaining that number or table. All the rest consists of people telling us what they have concluded on the basis of what they have assumed to be true.

I didn't understand why you said this the first time, and I don't understand why you repeat it now.

The following assumes that the "statement" is of the kind "Since X has value x, Y will be greater than y".

The answer is that there aren't any tables of numbers, and can't be, except for idealized conditions. The answer is not fixed for any given correlation. There are only methods. Richard, in the paper you redistributed a few days ago, demonstrated the equation-based method for the idealized situation in which the distribution is joint Gaussian, and provided numbers for particular levels of the cut. But you can probably use his methods and equations for other choices of the cut, such as "What is the probability that Y will be in the top quartile if X is above the median".

I say "probably" because I didn't reread the paper to check that all the equations are actually there. But I think they are. If not, it's simply a question of taking y' (the value of Y on the regression line when X has the value x), and seeing how many standard deviations y is above or below y', and looking in tables of the Gaussian distribution for your answer.

I gave you the most general method, which you can use for ANY distributional relationship. Remember that the correlation could be zero even when the two variables are functionally related, so that Y is uniquely determined by X. Here's an example: Y = X^2, when the tested X values are symmetrically distributed around zero. The correlation between X an Y is zero for that case, though you know Y exactly when you know X. The answer to your question "what is the probability that a statement about an individual that is based on group statistics will be incorrect" would be "Zero" for that case, even though the correlation is zero.

I'm not at all clear what more you could be asking for.

Martin

The second comment is similar: can anyone lay out the theoretical basis for libertarianism, representative democracy, rule of law, dog-eat-dog, or whatever system concept likes behind the political discussions? We're in a theory-based seminar here, so it seems appropriate to ask what theoretical bases there are for the various points of view.

You mean PCT-theoretic, I imagine. It's not an easy problem to address, though it is worthwhile to attempt.

One real problem that will arise is the same one that comes up in the medical treatment thread -- is it better to improve many people's lives at the cost of making other people's lives worse than they otherwise would have been (casting the question in the Conservative mode), or alternatively, is it better to make a few people's lives better at the cost of making many people's lives a bit worse (casting it in the Liberal mode). Either way you pose the question, the answer is a moral value judgment, not, I think, susceptible to scientific theoretical analysis, PCT or otherwise.

What is potentially susceptible to theory or simulation is whether certain arrangements are evolutionarily stable, and what kinds of dynamic changes are likely given certain starting assumptions. What, for example, might be the far future state of an isolated community with (un)limited resources in which most people choose to cooperate but some don't, and the cooperators allow the defectors to share in the benefits of cooperation without sanction (Samuel Sauders's preferred state, if I understand correctly). It might be possible to analyse and to simulate that oversimplification of a possible society, and likewise with the other concepts.

My replacement DSL model is supposed to be delivered tomorrow afternoon. I may come back to McDonald's tomorrow just to check up on things.

I hope she's pretty when she arrives!

Martin

4

[From Mike Acree (2007.08.02.1452 PDT)]

Bill Powers (2007.08.02.0834 MDT)–

can anyone lay out
the theoretical basis for libertarianism, representative democracy, rule of
law, dog-eat-dog, or whatever system concept likes >behind the political discussions?

I have come to the conclusion that what people are controlling for in
political disputes lies in the realm of morality (as they see it), rather than
politics per se, and ultimately in a sense of themselves as good (moral) people.
The indications of that are not especially subtle. I think the close
connection to a sense of self accounts for the extremely high gain exhibited in
these exchanges. I published an attempt at understanding such differences
a couple of years ago, and posted a link at that time, during a similar
exchange. As far as I can tell, only one person read the article (and
responded privately), but it continues to look relevant to me to the ongoing
discussion. I think I’ve achieved a fairly symmetric account of liberalism
and conservatism which holders of those views might endorse, and I extend the
same framework from politics into epistemology (just because I was intrigued
that epistemological anarchists tend to be political liberals and political
anarchists tend to be epistemological conservatives). The unsatisfying
part of the paper is my treatment of libertarianism. I voice some
criticism of libertarians, but the picture is not yet as theoretically coherent;
perhaps some of you can help. It is worth remarking in this context,
incidentally, that I have found the behavior of the recent libertarian, or
libertarianish, contributors to the Net to be an exception to my
stereotype. The posts of Saunders and Kennaway have been models of
civility in an area where even Bill has occasionally lost his exemplary cool.

http://libertyunbound.com/archive/2005_04/acree-daddy.html

Mike

5

[From Rick Marken (2007.08.02.1555)]

Mike Acree (2007.08.02.1452 PDT) --

The
unsatisfying part of the paper is my treatment of libertarianism. I voice
some criticism of libertarians, but the picture is not yet as theoretically
coherent; perhaps some of you can help.

I would start by trying to describe the phenomenon of libertarianism.
What is it? What does a libertarian want. Assuming what libertarians
want has to do with human behavior, I would then evaluate those wants
in terms of a model of human nature like, oh, say, Perceptual Control
Theory.

But why not start with describing what libertarianism _is_.

I'll try to do the same with liberalism when I get a chance. Of
course, it will have to be the Rick Marken version of liberalism; I
hope that will do.

Best

Rick

···

--
Richard S. Marken PhD
Lecturer in Psychology
UCLA
rsmarken@gmail.com

6

Re: Bill off net
[Martin Taylor 2007.08.02.20.54]

[From Mike Acree (2007.08.02.1452
PDT)]

http://libertyunbound.com/archive/2005_04/acree-daddy.html

It’s an interesting essay, putting a point of view I had not
considered. Perhaps because of that, it seems to me to present more of
a black-and-white cartoon tha a fully realized picture. Despite the
frequent aphorism that the only difference between conservatives and
socialists is what they want to prohibit (which is well presented in
the essay), I think there’s much more to it than that. One thing I do
agree with, however, is that most political disagreement stems from
differing views of morality. (Incidentally, I’m not sure I agree with
the presumed genesis of the differences in quasi-Freudian
psychology).

I don’t know where technical objections to libertarianism, that
it would probably be a self-defeating enterprise if taken seriously,
would enter your argument as to the reasons people reject it. Would it
be an affirmation of asceticism or its denial to suggest that if
increasing everyone’s ability to control their own perceptions is the
ultimate good, then libertarianism ought to be discarded, and
government by (coercively enforced) laws is desirable? That’s my
position, which I believe is supportable. Likewise, I support
increasing my own taxes, not for ascetic reasons, but because
analytically (and from observations of high-tax countries) I think
that if that happened I would have more, not less money to spend on
things I personally want (as opposed to the routine things like
doctors, food, transport, etc.). Is that an ascetic moral reason for
rejecting libertarianism?

I realize my analyses may be wrong, and there are large gaps of
unsupported assumption, but analysis is the start.

Martin

7

[From Bill Powers (2007.08.03.0710 MDT)]

Mike Acree (2007.08.02.1452 PDT) --

Every time I read an article by you, Mike, I envy your organization, knowledge, and mastery of the language.

In your article on "daddy" I think you can find some PCT ideas relating to theories behind libertarianism. Try scanning for programs/logic, principles, and system concepts, recalling that principles determine which programs or logic are used, and system concepts determine which principles are used. As you note in this article several times, logic doesn't seem sufficiently persuasive. According to my version of the hierarchy, that is because logic entails adoption of premises, and the choice of premises can be used to develop logical arguments that support any principle you wish. Furthermore, as you note, principles in the form of morality are not sufficiently persuasive, because the principles you put into effect depend on what system concept you have adopted and wish to support. So at the top of this heap are system concepts, and libertarianism is not the only system concept that exists at this level, even inside one libertarian. The tail of logic cannot wag the dog of system concepts.

How does one go about choosing a system concept, or choosing between competing system concepts? If I knew what the next level up is, I could answer that question, but I don't and can't. Basically I think we reorganize, shuffling the lower-level perceptions around -- morality and rationality -- looking for a better fit. When that still leaves us with a conflict, all we can do is hope that all this attention will lead to a new system concept to replace the old one or ones. So the arguments just go on, with no resolution.

That's the PCT theory we might apply. But when I asked for theories, I wasn't just thinking of PCT. Most libertarians, I assume, know nothing of PCT. So what theories are they using? What are they accepting as truths about human nature, about what causes behavior, about what sets goals and how goals relate to what we do? In short, how do they think people work? Or more to the point, what must necessarily be believed (at the levels of logic and principles) about how people work if libertarianism is to be defended and justified?

The same question can be asked of conservatives and liberals, Republicans and Democrats, Catholics and Protestants, and so on. At the moment I see no choice but to accept the system concepts behind these views as given, as chosen by the person for unfathomable reasons. All we can do is look at the structures of principles and rationality that have been fashioned to support them. Maybe if we look long enough, we will see something new. Maybe something will change.

Best,

Bill P.

8

Re: Bill off net

[From Mike Acree
(2007.08.03.1405 PDT)]

Many thanks to Martin and Bill for a
gratifying response.

Martin Taylor 2007.08.02.20.54–

I don’t know where technical objections to libertarianism, that it
would probably be a self-defeating enterprise if taken seriously, would enter
your >argument as to the reasons people reject it.

Technical objects to libertarianism are
obviously important (and they keep libertarians very busy arguing among
themselves), but were beyond the scope of that article, which was to try to understand
differences in political beliefs. I don’t think the reason most
people are liberals or conservatives is their technical objections to
libertarianism; in that respect you are very unusual in my experience. In
fact, that kind of coolly rational approach to political analysis is more
characteristic of libertarians, in my experience, and that is sometimes leveled
as a criticism. The empirical evidence on the viability of libertarianism
is generally so complicated as not to be convincing to anyone. Historical
examples of libertarian societies, for example—the U.S. from 1783 to 1789, or Iceland from
900 to 1300—are necessarily another time and place, and they didn’t
last, so they are easy to dismiss.

I support increasing my own taxes, not for ascetic reasons, but
because analytically (and from observations of high-tax countries) I think that
if that >happened I would have more, not less money to spend on things I
personally want (as opposed to the routine things like doctors, food,
transport, >etc.). Is that an ascetic moral reason for rejecting
libertarianism?

No, I accept your own statement that this is a purely data-driven
position, having nothing to do with moral asceticism, and that, if you saw the
data differently, you would be a libertarian. But there is a very
complicated process of data interpretation behind your claim, which would leave
a lot of room for dispute; and I don’t think that’s what’s
driving most people’s views.

Would it be an
affirmation of asceticism or its denial to suggest that if increasing
everyone’s ability to control their own perceptions is the ultimate good, >then libertarianism ought
to be discarded, and government by (coercively enforced) laws is desirable?
That’s my position, which I believe is >supportable.

This position carries not only empirical
but also logical difficulties, since the enforcement of laws is necessarily
reducing some people’s ability to control. I would certainly agree
that by constraining Bush’s ability to control (Bush as synecdoche; I’m
not sure he controls much of anything), we could allow millions of people much
greater control of their lives. But in this case we are specifically
controlling the controllers. It would be a large task to show that
constraining other people in general—constraining what substances they
were allowed to ingest, or what they were allowed to wear, or to whom they were
allowed to rent an apartment, or how much they were allowed to charge—led
in general, by some sort of social calculus, to greater freedom.

I think there’s one main reason why
libertarian societies would not succeed, in the foreseeable future, even if
they were established, and that is simply that that’s not what most
people want. They want a leader, they want someone in charge, even if
that will involve them in endless fights over whether the leader is one of
their team. If Dorothy Dinnerstein was too Freudian to be convincing on
that point, Julian Jaynes—or perhaps just looking around–would get you
to the same conclusion.

Sorry I will have difficulty continuing
this thread, if there’s anything to be followed up; I’m leaving
town for 2 weeks and will have infrequent access to e-mail. But thanks
again for your comments and questions.

Mike

9

[From Rick Marken (2007.08.03.1440)]

Mike Acree (2007.08.03.1405 PDT)

Historical examples of libertarian
societies, for example�the U.S. from 1783 to 1789, or Iceland from 900 to
1300�are necessarily another time and place, and they didn't last, so they
are easy to dismiss.

Why were these libertarian societies? What is a libertarian society?

Best

Rick

···

--
Richard S. Marken PhD
Lecturer in Psychology
UCLA
rsmarken@gmail.com

10

given a certain correlation,
what is the probability that a statement about an individual that
is based on group statistics will be
incorrect?
This is not a simple question.
If the correlation is perfect (1 or -1) then the answer is
everyone.
[From Bill Powers (2007.08.04.0835 MDT)]

Jeff Vancouver (2007.08.02.1545 EST) –

I think you misread the question. I was asking how often (for a given
correlation) a statement will prove to be incorrect regarding an
individual, when it is correct regarding the group data (i.e., the
“average person” in a study). For example, I once read a study
which concluded that “mothers hold their babies on the left
side,” the reason offered being that this was so the baby could hear
the heartbeat. But by no means all mothers held their babies on that
side. I forget the fraction of the mothers for whom the statement was
false, but it was pretty large. This was an example of a “fact”
that went down in the literature as true of “mothers,” when it
was true only of some mothers, and obviously false for the rest. How the
explanation applied to the rest of the mothers was not
discussed.

The second comment
is similar: can anyone lay out the theoretical basis for libertarianism,
representative democracy, rule of law, dog-eat-dog, or whatever system
concept likes behind the political discussions? We’re in a theory-based
seminar here, so it seems appropriate to ask what theoretical bases there
are for the various points of view.

Steven Pinder has a good discussion of these issues in his Blank
Slate book. In short, the answer seems twofold: first, individuals
have assumptions/theories that support their belief systems (this is
mostly what Pinder talks about), and second, that individuals do not seem
overly concerned with or capable of discerning internal consistency
within their belief system.

Is that a theory? It sounds more like a report of an observation of how
people behave. What I meant was the theory of how people work, the
principles that underlie the specific behaviors we observe. In PCT the
theory is that people are organized as negative feedback control systems
that act on the world to make their perceptions match internal reference
standards. In behaviorism, biology, neurology, and other fields the
theory is that stimuli act on the nervous system to alter the way inputs
cause output signals to activate the muscles. Libertarians appear to
believe that people are organized as control systems, but that they all
have certain perceptions and reference signals at higher levels such that
they will automatically create an ideal society if not subject to
coercion from others. I can’t tell from your brief allusions to
Pinder what kind of theory he adheres to.

The big issue in
the second statement is what does “overly” mean. It appears there is some
concern. It also appears there are individual differences in concern and
epistemology (i.e., critical analysis of information sought and
provided). An interesting PCT question is whether we desire consistency
and how we know (perceive) our degree of consistency. Social
psychologists seem to have accepted that we desire it (Leon Feshinger was
a big promoter of the concept). However, it seems highly likely that our
input function for creating the perception struggles or that the gain in
the system is low for some and/or some of the time. Indeed, I cannot
imagine doing anything else if all my resources had to be put to making
sure my belief system was consistent (true it did occupy a great deal of
my time when I was in high school – though girls still took precedent
even then). I just don’t think I could ever reduce that discrepancy
completely, though I do bump up gain or reference level from time to time
on that system (I think).

None of these points seem to have anything to do with theory – they
appear to be observational (or imagined) facts, though I doubt that any
one of them is true of every single individual studied. I would guess
that they derive from statistical analysis of studies, and that the
correlations leading to them are not exceptionally high. The underlying
theory would seem to be in the area of cognitive psychology, though
that’s only a guess. What is a “desire” for consistency? What,
exactly, is consistent with what (or not) when consistency is the issue?
Why would consistency be relevant? What is a “belief system?”
And so on.

Best,

Bill P.

11

[from Jeff Vancouver 2007.08.06.1037 EST)]

[From Bill Powers
(2007.08.04.0835 MDT)]

I think you misread the question. I was asking how often
(for a given correlation) a statement will prove to be incorrect regarding an
individual, when it is correct regarding the group data (i.e., the
“average person” in a study). For example, I once read a study which
concluded that “mothers hold their babies on the left side,” the
reason offered being that this was so the baby could hear the heartbeat. But by
no means all mothers held their babies on that side. I forget the fraction of
the mothers for whom the statement was false, but it was pretty large. This was
an example of a “fact” that went down in the literature as true of
“mothers,” when it was true only of some mothers, and obviously false
for the rest. How the explanation applied to the rest of the mothers was not
discussed.

This is a stupid conclusion that one is taught not
to make in elementary statistics and research methods classes. That said, it is
made all too frequently. My point is that a correlation does not provide the
information about frequency (though some kinds can, e.g., logistic regression).

Steven Pinder has a good discussion of these issues in his Blank
Slate book. In short, the answer seems twofold: first, individuals have
assumptions/theories that support their belief systems (this is mostly what
Pinder talks about), and second, that individuals do not seem overly concerned
with or capable of discerning internal consistency within their belief system.

Is that a theory? It sounds more like a report of an observation of how people
behave. What I meant was the theory of how people work, the principles that
underlie the specific behaviors we observe.

You misunderstand me. By theory I mean naïve to sophisticated
imaginations about the way the world and people in it work. Everyone has them
and they influence belief systems (indeed, they are part of them). Meanwhile Pinder
is not promoting a particular theory. He was promoting the idea that nature
(not just nurture) needs to be taken seriously within theories. But that is
irrelevant to my point.

The big issue in the second statement is what does
“overly” mean. It appears there is some concern. It also appears
there are individual differences in concern and epistemology (i.e., critical
analysis of information sought and provided). An interesting PCT question is
whether we desire consistency and how we know (perceive) our degree of
consistency. Social psychologists seem to have accepted that we desire it (Leon
Feshinger was a big promoter of the concept). However, it seems highly likely
that our input function for creating the perception struggles or that the gain
in the system is low for some and/or some of the time. Indeed, I cannot imagine
doing anything else if all my resources had to be put to making sure my belief
system was consistent (true it did occupy a great deal of my time when I was in
high school – though girls still took precedent even then). I just
don’t think I could ever reduce that discrepancy completely, though I do
bump up gain or reference level from time to time on that system (I think).

None of these points seem to have anything to do with theory – they appear to
be observational (or imagined) facts, though I doubt that any one of them is
true of every single individual studied. I would guess that they derive from
statistical analysis of studies, and that the correlations leading to them are
not exceptionally high. The underlying theory would seem to be in the area of
cognitive psychology, though that’s only a guess. What is a “desire”
for consistency? What, exactly, is consistent with what (or not) when
consistency is the issue? Why would consistency be relevant? What is a
“belief system?” And so on.

These points have much to do
with theory. Take for example your case. You have beliefs, based on
sophisticate epistemology thankfully, about human as feedback control systems.
Some aspects of those beliefs you are comfortable with and some you are not so
sure about. You apply modeling because you believe it reveals inconsistencies
or allows you to confirm/identify predictions of your theories. If the models don’t
do that, you question the models or the aspect of the theory, because you do
not like the inconsistency (perhaps, this is only a theory of course). Moreover,
your questions are like mine (though not all of them). They are the hypothetical
“is consistency a controlled variable” (for anyone; everyone but
with different parameters).

Jeff V.

12

[From Rick Marken (2007.08.06.1000)]

Jeff Vancouver 2007.08.06.1037 EST)

Bill Powers (2007.08.04.0835 MDT)--

I think you misread the question. I was asking how often (for a given
correlation) a statement will prove to be incorrect regarding an individual,
when it is correct regarding the group data (i.e., the "average person" in a
study). For example, I once read a study which concluded that "mothers hold
their babies on the left side,"... This was an example of a
"fact" that went down in the literature as true of "mothers,"

This is a stupid conclusion that one is taught not to make in elementary
statistics and research methods classes.

I have taught both statistics and research methods several times now
since "retiring" and I don't recall ever seeing a textbook that
included a caveat against concluding that the group results apply to
any individual in the population. In fact I have never seen a
psychological report that included this caveat in its conclusion.
Every report with which I am familiar implies that the group result
applies to all people (just as in Bill's example, where the conclusion
is that "mother's hold their babies on the left side" based on data
that probably show that somewhat more than 50% of mothers hold their
babies on the left side somewhat more than 50% of the time). Results
of psychological research on groups (which is nearly all psychological
research) are reported this way (eg. "Viewing violence causes violent
behavior", the implication being that it causes all people to behave
violently) or as a "tendency" (eg. "People tend to overestimate the
probability of rare events" inplying that all people do this but that
it doesn't occur every time you observe a person estimating the
probability of a rare event because of "sampling error").

That said, it is made all too frequently.

My guess is that it would be hard to find a report of research on
groups of subjects in a psychological journal that does not make this
mistake. But, then, people do tend to overestimate the probability of
rare events;-)

Best

Rick

···

--
Richard S. Marken PhD
Lecturer in Psychology
UCLA
rsmarken@gmail.com

13

[from Jeff Vancouver (2007.08.08.1300 EST)]

[From Rick Marken (2007.08.06.1000)]

I have taught both statistics and research methods several times now
since "retiring" and I don't recall ever seeing a textbook that
included a caveat against concluding that the group results apply to
any individual in the population.

Hi Rick,

First, I assume you meant "sample" not "population." There is almost always
a chapter devoted to distinguishing descriptive from inferential statistics
(and the fact they are not isomorphic [i.e., inferential!]).

However, even if sample is used, I must say that I have taught both stats
and research methods as well and recall seeing this several times. I believe
there is a classic article on it, but I cannot remember the author (but I
would guess it was Paul Meehl; could be Rogosa). I now teach a graduate
level course in research methods and often ask if they would interpret a
mean difference this way (or a correlation as absolute). They are usually
clear that such an interpretation would be a mistake. In other words, they
learned it as undergraduates (though I suspect some students have forgot or
never learned it). Moreover, it follows from understanding of variance
(e.g., a picture of two overlapping distributions is often used in these
books) or descriptions of residuals. I would be surprised if a methods
journal was impressed by another trying to make this point (but it is your
time). Indeed, at some level it is offensive (implying us researchers are
that stupid).

I also found an article that provides a method for answering Bill's question
when the independent variable is a dichotomous variable
(intervention/placebo), but the dependent variable is continuous. It looks
like they basically dichotomizes the continuous variable so logistic
approaches could be use (which provide odds ratios, etc.).

http://findarticles.com/p/articles/mi_m0999/is_7132_316/ai_53499438

FYI, they starts from the same assumption as yours (that researchers
misinterpret a mean value as applying to everyone in the group), but for low
(and not statistically significant) differences. They suggest that hidden in
that number is that often some minority are helped and that could be
important (particularly if cost is low).

Jeff V.

14

[From Rick Marken (2007.08.08.1045)]

Jeff Vancouver (2007.08.08.1300 EST) --

> Rick Marken (2007.08.06.1000)]
>
> I have taught both statistics and research methods several times now
> since "retiring" and I don't recall ever seeing a textbook that
> included a caveat against concluding that the group results apply to
> any individual in the population.

Hi Rick,

First, I assume you meant "sample" not "population."

No, I actually meant population. When you do a study on a sample you
use the results to infer what is true of the population. So if the
sample result is, say, that the average aggression score is higher for
the group that observed an aggressive model than for the group that
didn't, then your statistics allow you to decide whether this result
is true of the population. Assuming that the you do decide this (the
results is "significant"), then what I was saying was that research
methods text should then caveat against concluding that this result
applies to any individual in the population. The conclusion based on
the group (sample) data applies only to another group (population).

However, even if sample is used, I must say that I have taught both stats
and research methods as well and recall seeing this several times. I believe
there is a classic article on it, but I cannot remember the author (but I
would guess it was Paul Meehl; could be Rogosa).

Yes. I think I recall a paper on "idiographic" (individual) vs
"nomothethic" (group) results. But it hasn't really caught on in
psychology, has it. The idiographic approach is non-statistical. How
much research in psychology is non-statistical? Some Skinnerian stuff,
peraps. But almost all the research reported in the major
psychological journals is group data and the results are regularly
reported as though they applied to individuals. That is, they are
reported the way Bill said the results of the "baby side" study are
reported: "mothers carry their baby's on the left side". It's good
that you give the caveat about the lack of applicability of group
results to individuals. But I don't think it has really taken hold in
the field.

I now teach a graduate
level course in research methods and often ask if they would interpret a
mean difference this way (or a correlation as absolute). They are usually
clear that such an interpretation would be a mistake. In other words, they
learned it as undergraduates (though I suspect some students have forgot or
never learned it). Moreover, it follows from understanding of variance
(e.g., a picture of two overlapping distributions is often used in these
books) or descriptions of residuals. I would be surprised if a methods
journal was impressed by another trying to make this point (but it is your
time). Indeed, at some level it is offensive (implying us researchers are
that stupid).

Well, when I stop seeing group level research done to test theories of
individual behavior then I'll stop being offensive. I certainly don't
mean to imply that researchers who do this are stupid, by the way.
Like religious fundamentalists who have been to college, you have to
be very smart, I think, to maintain ridiculous dogma in the face of
obvious fact.

Best

Rick

···

--
Richard S. Marken PhD
Lecturer in Psychology
UCLA
rsmarken@gmail.com

15

Yes. I think I recall a paper on
“idiographic” (individual) vs

“nomothethic” (group) results.
[From Bill Powers (2007.08.08.1350 MDT)]

Rick Marken (2007.08.08.1045)] –

I’ve heard about that, too, believe it or not. Such fancy words! I often
wondered why it wasn’t “idiothetic” or “nomographic”,
to make the construction more parallel. What is “self-drawing?”
Or “name-feeling?”

Well, when I stop
seeing group level research done to test theories of

individual behavior then I’ll stop being offensive. I certainly
don’t

mean to imply that researchers who do this are stupid, by the way.

Like religious fundamentalists who have been to college, you have to

be very smart, I think, to maintain ridiculous dogma in the face of

obvious fact.

How about getting out one of those journals and going
through a few issues of it? Maybe Jeff can suggest one. I really think
you’re right – it’s a rare paper that doesn’t apply group data to
individuals – but we need some concrete cases before us to focus the
discussion.

Best,

Bill P.

16

[from Jeff Vancouver (2007.08.08.1616 EST)]

[From Rick Marken (2007.08.08.1045)]

No, I actually meant population. When you do a study on a sample you
use the results to infer what is true of the population. So if the
sample result is, say, that the average aggression score is higher for
the group that observed an aggressive model than for the group that
didn't, then your statistics allow you to decide whether this result
is true of the population. Assuming that the you do decide this (the
results is "significant"), then what I was saying was that research
methods text should then caveat against concluding that this result
applies to any individual in the population. The conclusion based on
the group (sample) data applies only to another group (population).

Yeah, this would be a twice removed inference (based on average in a
population). BTW, I see "on average" used all the time in research. The
point of that phrase is to remind the reader that the findings are referring
to a group mean, not all the individuals in the group. Then statistics
describes probabilities of seeing the effect repeated on others in the
population. Finally, there is often explicit discussion of generalizability
particularly in applied studies.

Yes. I think I recall a paper on "idiographic" (individual) vs
"nomothethic" (group) results. But it hasn't really caught on in
psychology, has it. The idiographic approach is non-statistical. How
much research in psychology is non-statistical? Some Skinnerian stuff,
peraps. But almost all the research reported in the major
psychological journals is group data and the results are regularly
reported as though they applied to individuals. That is, they are
reported the way Bill said the results of the "baby side" study are
reported: "mothers carry their baby's on the left side". It's good
that you give the caveat about the lack of applicability of group
results to individuals. But I don't think it has really taken hold in
the field.

It has in a number of places. In particular, multilevel data analysis
techniques have been a boon to looking at differences in individuals
processes (e.g., individual growth curves). See for example Singer and
Willet (2003).

> I now teach a graduate
> level course in research methods and often ask if they would
interpret a
> mean difference this way (or a correlation as absolute). They are
usually
> clear that such an interpretation would be a mistake. In other words,
they
> learned it as undergraduates (though I suspect some students have
forgot or
> never learned it). Moreover, it follows from understanding of
variance
> (e.g., a picture of two overlapping distributions is often used in
these
> books) or descriptions of residuals. I would be surprised if a
methods
> journal was impressed by another trying to make this point (but it is
your
> time). Indeed, at some level it is offensive (implying us researchers
are
> that stupid).

Well, when I stop seeing group level research done to test theories of
individual behavior then I'll stop being offensive. I certainly don't
mean to imply that researchers who do this are stupid, by the way.
Like religious fundamentalists who have been to college, you have to
be very smart, I think, to maintain ridiculous dogma in the face of
obvious fact.

But I think you missed Bill's point. Group research can be used to examine
individual models.

Jeff V.

17

But I think you missed Bill’s
point. Group research can be used to examine

individual models.
[From Bill Powers (2007.08.08.1540 MDT)]
Jeff Vancouver (2007.08.08.1616 EST) –
In the data Rick is examining, it turns out that you have to aggregate
the group data into rather large bins (a quarter or a third of the range
of the independent variable) to get high accuracy of prediction. In the
present case this means you’re making predictions for 30 to 40 countries
in a bunch, not individual countries.
Rick, if you use Martin’s CIA spreadsheet data, there are two rows that
have incomplete data – 113 (Cyprus area) and 182 (Falkland Islands). The
correlations still work all right because they don’t use the missing
entries. The correlation of Log10(GDP) vs Log10(IM) is a shade under
0.82. It will be interesting to see how much worse the “fine
slicing” problem becomes at this lower correlation.
Jeff, my present impression is that to make predictions for individuals
with a 90% chance of being correct, the initial correlation behind the
regression line has to be well into the 0.9s, maybe the 0.99s. But we’re
still working on that. Do you have any data we could look at?
Actually group data should never be used for testing models. The way it
should be done is to match the parameters of a model for the best fit to
each individual’s data first, and only then look at the group data about
the parameters. It really makes little sense, even then, to speak of
group characteristics, like “average loop gain” and
“average reference level.” You can certainly calculate such
numbers, but what do they mean? Perhaps there would be some point in
looking for unusual deviations from the ordinary distribution of these
parameters in a population, for example when looking for pathology. But
for predicting individual behavior, you need a very good model fit
before the average model will be of much use.

Best,

Bill P.