FacebookTwitter
Hatrack River Forum Post New Topic  Post A Reply
my profile login | register | search | faq | forum home

  next oldest topic   next newest topic
» Hatrack River Forum » Active Forums » Books, Films, Food and Culture » LDS Community and marriage (Page 8)

  This topic comprises 9 pages: 1  2  3  4  5  6  7  8  9   
Author Topic: LDS Community and marriage
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
Church-going rates are a reasonable measure of religion in the country.
And, as you'll observe from the figures in my link, it is just as strongly correlated with homicide etc as any of the other variables used. You could drop the absolute belief, prayer, and belief in evolution, and the conclusion would still be the same. This being so, will you kindly withdraw this objection to the link?

quote:
Comparing states or countries in that way involves way too many confounding variables that you are not controlling for.
I will ask you the same question I asked CT: Does this study make it, in your estimation, more likely or less likely that religion leads to homicides?
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
Let's try some Bayes. Suppose you did not know about the study I just linked. Suppose further that you had some probability estimate of the proposition "Religious belief makes people more likely to murder" - say, 5%. Then I tell you about the study. Does your estimate increase, decrease, or stay the same? Notice that I do not ask about quantities, since I made up the 5% anyway, but about directions.

[edited for snark; my apologies, out of line]

The Discussion section near the end of most relevant journal articles usually includes some bit about this issue, often at great and tiresome length.

"With qualifications" does not equal "impossible to make any deductions whatsoever," and it never has. But the qualifications are what makes for rigor and intellectual integrity, and they are what help us balance more carefully as we try to move from research to policy. We have to make policy, but our understanding is never set in stone, so we have to make policy knowing it may change and with the acknowledged willingness to anticipate and seek out such changes, and make whatever corrections to policy are indicated, if so.

[ May 15, 2009, 01:53 PM: Message edited by: ClaudiaTherese ]

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
I will ask you the same question I asked CT: Does this study make it, in your estimation, more likely or less likely that religion leads to homicides?

[edited for snark; my apologies, out of line]

Your study would not lead me to conclude anything one way or the other as regards likelihood. It would make me curious about certain directions, generate additional and potentially fruitful hypotheses, and likely be the nitus of several interesting phone calls, though.

But conclusions as to meaning? None. Possibilities and interesting areas to look further? Yes.

What makes them not be firm conclusions is the required qualifications, and it is the qualifications that raise the further questions.

[ May 15, 2009, 01:55 PM: Message edited by: ClaudiaTherese ]

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
And, as you'll observe from the figures in my link, it is just as strongly correlated with homicide etc as any of the other variables used.
No its not. You are completely harpooning your own credibility as a scientists by making such statements. Look at the graph!! The only reason there is any correlation between religiosity and murder is because the US and Poland happen to be high in both factors. Leave them off and there is no correlation at all. If you look at frequent church attendance, the US and Poland are off scale in the murder rate but in the mid range for church attendance which completely destroys what ever correlation existed.

I'm in the middle of marking student research papers and if one of my students made a conclusion like that from this data set, I'd fail them. If you are actually a Ph.D student in physics as you've claimed, you must be doing solely theoretical work because your data analysis skills are really atrocious.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Does this study make it, in your estimation, more likely or less likely that religion leads to homicides?
Based on the data presented in this study, I do not think it make a significant contribution to the debate. All I can glean from it is that most societal factors in the US look more like developing countries than they do other developed countries, which is curious but not evidence that religiosity is the cause.
Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
I get what you are saying, King of Men (or at least, I'm pretty sure I do). I'll also confess to a healthy skepticism that committing to a religion makes one a better person, helps one to be a better person, or even is an aid to preventing one from becoming a worse person (whatever the definitions may be).

I just honestly don't take anything away from the study other than interesting questions, as, quite frankly, the alternative explanations are just as viable [and the potential confounders play havoc so readily].

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by The Rabbit:
[QB] Even if you have prior reason to believe that religion has a 5% probability of making people more likely to murder, that wouldn't make these correlations significant at any reasonable standard of evidence.

Let us run some dang numbers. I am picking a figure at random; if you like you can pick a different one and run the same analysis. Figure four, middle of the second column, "Atheists and agnostics" versus "Under-five mortality/1000 births". Reading off from the figure:

P: (7, 9)
I: (7, 7)
T: (10, 6)
L: (15, 5)
R: (16, 5)
S: (16, 6)
C: (19, 6)
Z: (21, 6)
N: (25, 4) (Go Norway!)
J: (31, 4)
W: (35, 4)
A: (25, 5)
E: (26, 6)
H: (30, 5)
D: (31, 5)
G: (34, 5)
F: (39, 5)

Notice that I left out the US. The correlation coefficient is -71%. Seventy-one percent, dude! And this is 'trash'? What sort of data do you work with, then, that has so much stronger correlations?

Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
Tresopax
Member
Member # 1063

 - posted      Profile for Tresopax           Edit/Delete Post   Reply With Quote 
quote:
And, as you'll observe from the figures in my link, it is just as strongly correlated with homicide etc as any of the other variables used. You could drop the absolute belief, prayer, and belief in evolution, and the conclusion would still be the same. This being so, will you kindly withdraw this objection to the link?
Well, if you look at only the charts comparing to "church-going", the only correlation that I see clearly is to abortion, especially once you take out the U.S. And arguably infant deaths. That to me suggests there could be some connection between religiousness and teen pregnancy, but I suspect poverty or culture are the confounding variables causing it.

quote:
Does this study make it, in your estimation, more likely or less likely that religion leads to homicides?
I'd say more likely. But then again, a random guy on the street telling me religion leads to homicides also makes it more likely to be true.
Posts: 8120 | Registered: Jul 2000  |  IP: Logged | Report this post to a Moderator
Tresopax
Member
Member # 1063

 - posted      Profile for Tresopax           Edit/Delete Post   Reply With Quote 
quote:
Notice that I left out the US. The correlation coefficient is -71%. Seventy-one percent, dude! And this is 'trash'? What sort of data do you work with, then, that has so much stronger correlations?
Again, I think you could find a similar correlation between average temperature and religion if you compared U.S. states.
Posts: 8120 | Registered: Jul 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
Your study would not lead me to conclude anything one way or the other as regards likelihood.
I am sure you can do better than this. You apparently agree that this data is not utterly irrelevant. It follows that it must have some sort of effect on your probability estimate. Which direction? Up or down? It can be a very tiny change if you like, but unless you are willing to state (with Rabbit) that these data are completely irrelevant, your estimate has to go one way or the other. Please choose.

Rabbit, you posted while I was working out that correlation coefficient, and I believe you have now crossed the line into personal attack. You probably did not intend to do so. Perhaps you would like to back down?

Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
katharina
Member
Member # 827

 - posted      Profile for katharina   Email katharina         Edit/Delete Post   Reply With Quote 
She's not wrong, King of Men. Statistical analysis does not support your conclusions.
Posts: 26076 | Registered: Mar 2000  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
It follows that it must have some sort of effect on your probability estimate. Which direction? Up or down?

My apologies, but I am constrained by the rigor of my training. [Smile] I do not make assertions without possibility of tenable substantiation.

No matter how often someone may ask me to do so.

---

Form the article:

quote:
Regression analyses were not executed because of the high variability of degree of correlation, because potential causal factors for rates of societal function are complex, and because it is not the purpose of this initial study to definitively demonstrate a causal link between religion and social conditions. Nor were multivariate analyses used because they risk manipulating the data to produce errant or desired results,<5> and because the fairly consistent characteristics of the sample automatically minimizes the need to correct for external multiple factors (see further discussion below). Therefore correlations of raw data are used for this initial examination.
The qualifications already acknowledged by the [author poses] limitations on any conclusions that can be drawn. As far as I can see, there is no attempt to take into account the potential confounding of socioeconomic status, either in aggregate or in terms of spread between highest and lowest in a country.

[SES is the most cliched but also most common confounder in the business. It is never irrelevant enough to disregard without mention in this sort of study.]

That isn't fancy laboring over the data, but rather basic, solid, analytic technique.

All this raw data can be good for is purest speculation, nothing more. And if the data were to be presented without proper qualifications extensively discussed, it would be shoddy work that lessens the understanding about the topic, not furthers it.

Are there any responses to this article in the literature? Other articles that cite it, positively or negatively?

[ May 15, 2009, 02:29 PM: Message edited by: ClaudiaTherese ]

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
MrSquicky
Member
Member # 1802

 - posted      Profile for MrSquicky   Email MrSquicky         Edit/Delete Post   Reply With Quote 
Man, it's like reading creationists attack evolution.

Step 1: Start with some information
Step 2: Filter it through a lack of understanding of how science works
Step 3: Declare it shows whatever conclusion you started out with
Step 4: ...
Step 5: Profit!

Posts: 10177 | Registered: Apr 2001  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
The author of the study King of Men referenced, ""Cross-National Correlations of Quantifiable Societal Health with Popular Religiosity and Secularism in the Prosperous Democracies: A First Look"," is Gregory S. Paul (link is to Wikipedia), described as "a freelance paleontologist, author and illustrator."

The Wikipedia article list several areas of criticism regarding that particular paper. I'm excerpting Wikipedia below for anyone who wants to pursue primary sources of the criticisms referenced there. I won't, but I'd be happy to read along on any discussion that follows. As it is, though, other areas of my life are pressing.

quote:
Religion
See also: Morality#Religiosity and morality

Paul authored a paper in 2005 entitled "Cross-National Correlations of Quantifiable Societal Health with Popular Religiosity and Secularism in the Prosperous Democracies: A First Look".[2] He states in the introduction that the paper is "not an attempt to present a definitive study that establishes cause versus effect between religiosity, secularism and societal health".[3] This paper has been criticized on statistical grounds, for conceptual ambiguity , its indirect measure of "religiosity" (the author's term) and its "chi-by-eye" interpretation of scatterplots rather than quantified measures. Summing up in a published article in the same journal, Moreno-Riaño, Smith, and Mach from Cedarville University wrote that "[Paul's] methodological problems do not allow for any conclusive statement to be advanced regarding the various hypotheses Paul seeks to demonstrate or falsify."[4] [bold added for emphasis, and it summarizes well my opinion on the matter --CT] At the time the paper was published, Paul announced plans to write a book on the subject, claiming that the findings are strong enough to justify further study.[5]

Gary F. Jensen of Vanderbilt University is one of the scientists who criticizes the methods used by Paul, including that "Paul’s analysis generates the 'desired results' by selectively choosing the set of social problems to include to highlight the negative consequences of religion". In a response [6] to the study by Paul, he builds on and refines Paul's analysis. His conclusion, that focus only in the crime of homicide, is that there is a correlation (and perhaps a causal relationship) of higher homicide rates, not with Christianity, but with dualistic Christian beliefs, something Jensen defines as the strong belief in all of the following : God, heaven, devil and hell. Excerpt: "A multiple regression analysis reveals a complex relationship with some dimensions of religiosity encouraging homicide and other dimensions discouraging it."

---

Added: Wikipedia listing of relevant references and footnotes:

quote:
Footnotes

1. ^ Paul, Gregory S. (1988). Predatory Dinosaurs of the World. New York: Simon & Schuster. ISBN 0-671-61946-2.
2. ^ Gregory S. Paul (2005), "Cross-National Correlations of Quantifiable Societal Health with Popular Religiosity and Secularism in the Prosperous Democracies". Journal of Religion and Society. Vol. 7. [1]
3. ^ Paul 2005, p. 2.
4. ^ "Religiosity, Secularism, and Social Health". [2]. 2006. http://moses.creighton.edu/JRS/2006/2006-1.html. Retrieved on 2007-05-27.
5. ^ "Religion and social problems". MSNBC. 2005. http://www.msnbc.msn.com/id/21134540/vp/10364362#10364362. Retrieved on 2009-02-19.
6. ^ Jensen 2006.

References

* Jensen, G. F. (2006). "Religious Cosmologies and Homicide Rates among Nations" (PDF). Journal of Religion & Society 8: 1–14. ISSN 1522-5658. http://moses.creighton.edu/JRS/pdf/2006-7.pdf. Retrieved on 2007-04-07.
* Paul, G. S. (2005). "Cross-National Correlations of Quantifiable Societal Health with Popularity Religiosity and Secularism in the Prosperous Democracies" (PDF). Journal of Religion & Society 7: 1–17. ISSN 1522-5658. http://moses.creighton.edu/JRS/pdf/2005-11.pdf. Retrieved on 2007-04-07.
* Paul, G.S. (2002). Dinosaurs of the Air: The Evolution and Loss of Flight in Dinosaurs and Birds. Baltimore: Johns Hopkins University Press.



[ May 15, 2009, 02:43 PM: Message edited by: ClaudiaTherese ]

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
advice for robots
Member
Member # 2544

 - posted      Profile for advice for robots           Edit/Delete Post   Reply With Quote 
My view of humanity is apparently very rosy.

I'll retract my previous statement about men and using authority for good. It is a fairly safe assumption that they will use it for wrong if they get it. Still, it is my privilege to know many who don't.

Posts: 5957 | Registered: Oct 2001  |  IP: Logged | Report this post to a Moderator
TomDavidson
Member
Member # 124

 - posted      Profile for TomDavidson   Email TomDavidson         Edit/Delete Post   Reply With Quote 
quote:
Again, I think you could find a similar correlation between average temperature and religion if you compared U.S. states.
Well, since only stupid people live in hot climates... [Wink]

*keeps tongue firmly in cheek*

Posts: 37419 | Registered: May 1999  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
quote:
Originally posted by The Rabbit:
[QB] Even if you have prior reason to believe that religion has a 5% probability of making people more likely to murder, that wouldn't make these correlations significant at any reasonable standard of evidence.

Let us run some dang numbers. I am picking a figure at random; if you like you can pick a different one and run the same analysis. Figure four, middle of the second column, "Atheists and agnostics" versus "Under-five mortality/1000 births". Reading off from the figure:

P: (7, 9)
I: (7, 7)
T: (10, 6)
L: (15, 5)
R: (16, 5)
S: (16, 6)
C: (19, 6)
Z: (21, 6)
N: (25, 4) (Go Norway!)
J: (31, 4)
W: (35, 4)
A: (25, 5)
E: (26, 6)
H: (30, 5)
D: (31, 5)
G: (34, 5)
F: (39, 5)

Notice that I left out the US. The correlation coefficient is -71%. Seventy-one percent, dude! And this is 'trash'? What sort of data do you work with, then, that has so much stronger correlations?

KOM, I don't know if you are being dishonest or just nor reading thoroughly but figure 4 (religiosity vs early childhood mortality) is one of the cases where I already admitted the data is correlated.

Instead, lets pick the graph of church attendence vs murder, which you specifically claimed showed a correlation. Reading the data off the chart (if you want I can post the numbers I read) I get a correlation coeficient of 0.0356 with all countries included. If you exclude the US, the correlation coefficient drops to 0.0125. If you exclude both the use and Poland, you actually get a very slight anti-correlation with an R2 of 0.005.

Does that make my point clear?

The human brain is programmed to see patterns even when none exist. This is particularly true when one already has preconceived notions about what the data says. This is why statistical analysis is so important and why this paper is a load of crap.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
I gave numbers. So far katharina and Rabbit have responded with assertions of "You're wrong" and Squicky and Rabbit have added insults. Now, it's possible that I was unlucky in my random pick and got the only large correlation in the whole lot, which would make the significance drop considerably. If that's so, then fine, I was mistaken and the study does not actually support my claim. Nobody has even attempted to argue reasonably in this direction, unless you count Rabbit's "Your data analysis skills are really atrocious". Is this really the standard of argument you people want to hold yourself to?

quote:
My apologies, but I am constrained by the rigor of my training. [Smile] I do not make assertions without possibility of tenable substantiation.
Your rigorous training does not appear to include Bayesian methods. Is it possible that you misunderstood my question? I do not ask for an estimate of X in "There is X% probability that religion will cause a homicide." I ask for a degree of belief in "Religion increases homicides", as in, if an all-knowing god offers you a bet on this statement, and will reveal the answer, what odds do you want? Unless you are asserting either that these data do not bear on that question at all, or else that they are equally likely to appear whether the statement is true or false, then your requested odds must change. Right?
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
All right, the homicide graph does not show a correlation. Was that such a hard argument to make? What's with the insults?
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
MrSquicky
Member
Member # 1802

 - posted      Profile for MrSquicky   Email MrSquicky         Edit/Delete Post   Reply With Quote 
KOM,
The sources of error make it so that to a real scientist, for the purposes you are trying to put it, the valid information from that data is non-existent. This is another case where your ignorance of basic scientific epistemology (and heck, statistical inference) is coming through.

[ May 15, 2009, 03:17 PM: Message edited by: MrSquicky ]

Posts: 10177 | Registered: Apr 2001  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
And then there is this follow-up paper by a different author.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
ClaudiaTherese
Member
Member # 923

 - posted      Profile for ClaudiaTherese           Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
quote:
My apologies, but I am constrained by the rigor of my training. [Smile] I do not make assertions without possibility of tenable substantiation.
Your rigorous training does not appear to include Bayesian methods. Is it possible that you misunderstood my question?
"No matter how often someone may ask me to do so."

quote:
Summing up in a published article in the same journal, Moreno-Riaño, Smith, and Mach from Cedarville University wrote that "[Paul's] methodological problems do not allow for any conclusive statement to be advanced regarding the various hypotheses Paul seeks to demonstrate or falsify.
Or those you do. [As MrSquicky puts it so succinctly, "the valid information from that data is non-existent." Indeed, it is.]

I am sorry, but I have nothing further to say on the matter, and I will have nothing further to say.

Some questions are not meant for answering, but rather for allowing to be left hanging, isolated against the clear blue sky, swinging gently in the breeze. (Unfortunately for the impatient and importunate, this is one of them.)

Posts: 14017 | Registered: May 2000  |  IP: Logged | Report this post to a Moderator
MrSquicky
Member
Member # 1802

 - posted      Profile for MrSquicky   Email MrSquicky         Edit/Delete Post   Reply With Quote 
quote:
I'll retract my previous statement about men and using authority for good. It is a fairly safe assumption that they will use it for wrong if they get it. Still, it is my privilege to know many who don't.
I've known many who would use it correctly as well, but I've known many more who would abuse it. And historically, the (often religiously justified) idea that men hold authority over women has been put to uses that are very different from what you are saying it should be.

In fact, forget historically. This is going on right now. I'm willing to bet that this is even true for a not insignificant number of people in your religion.

---

edit: I'm not sure how to get my concern across here, so this may be clumsy. I think people are often more concerned with justifying the potentially problematic things they believe instead of guarding against the problems that very forseeably develop from people believing them. Also, this very instinct to justify often abets the furtherance of these problems and abuses.

Posts: 10177 | Registered: Apr 2001  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
Oh, and by the way KOM. If you are talking about the percent of variance in a data set that is described by a linear correlation, you should be reporting R squared and R. Which for graph #4 turns out to be 50.6% not 71%.
Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
MrSquicky
Member
Member # 1802

 - posted      Profile for MrSquicky   Email MrSquicky         Edit/Delete Post   Reply With Quote 
KOM,
You, I suspect consciously, present yourself who understands how science works, but you don't. I note this when it becomes very apparent so that people who are not as acquainted with what science actually is don't see your mistaken ideas presented as valid without challenge.

Posts: 10177 | Registered: Apr 2001  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
Evidentally one of my earlier post got lost. Here is roughly what I said some where around the bottom of the last page

The authors of this paper claim,

quote:
In general, higher rates of belief in and worship of a creator correlate with higher rates of homicide, juvenile and early adult mortality, STD infection rates, teen pregnancy, and abortion in the prosperous democracies (figure 1-9).
Of those 6 claims, only 2 (early childhood mortality and abortion rates) are evident in the data they present. The remaining claims are not support even by the data the authors present. With those two exceptions, any correlations in the data are not statistically significant whether or not the US is included in the data.

It is worth noting that early childhood mortality and abortion rates have been shown to be strongly correlated to poverty and that the correlation between income disparity in the countries studied and childhood mortality and abortions is much stronger than the correlation with religiosity.

It is an interesting note that there appears to be a correlation between religiosity in these countries and income disparity and I think the reason for that is worth discussing.

The paper in question is not.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Rabbit, you posted while I was working out that correlation coefficient, and I believe you have now crossed the line into personal attack. You probably did not intend to do so. Perhaps you would like to back down?
My personal attack was specifically regarding your statement that murder rates as strongly correlated to church attendance as other factors of religiosity. If you had said childhood mortality was as strongly correlated to church attendance as other factors, I would have agreed. If you don't wish to be ridiculed for bad science, don't present yourself as an accomplished scientist and then make claims that are scientifically ridiculous.

No I don't plan to back down.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by The Rabbit:
Oh, and by the way KOM. If you are talking about the percent of variance in a data set that is described by a linear correlation, you should be reporting R squared and R. Which for graph #4 turns out to be 50.6% not 71%.

I'm not doing a fit, I'm just calculating the correlation coefficient. Since the information content is the same, saying that or the other 'should' be used is purely conventional. If you want to argue that R^2 is more often used in social sciences, that's fine; I wouldn't know. But there's no 'should' about it.

If your lost post had not been lost, this discussion would have been a lot more pleasant. Perhaps you would like to retract your accusation of dishonesty and/or not reading your posts, now?

Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
My personal attack was specifically regarding your statement that murder rates as strongly correlated to church attendance as other factors of religiosity.
Well, I admit I was looking at Figure 4 when I said that, so the phrase "homicide etc" was rather unfortunate, careless, and mistaken. A simple "Hang on, you'd better check that" would have done the job, though.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
advice for robots
Member
Member # 2544

 - posted      Profile for advice for robots           Edit/Delete Post   Reply With Quote 
Squicky: It unfortunately is true for many in my religion, who apparently do not understand what the authority is and how it may be exercised.

If they are using it to secure power and control for themselves, that is a sure sign they are misusing their position and therefore will have that authority taken away and be left to their own devices. Abuse of one's position in the marriage and family is not a characteristic of priesthood authority.

Even if it can be misunderstood and the position of a priesthood holder misused, however, we believe priesthood authority is necessary and does much good that could not be done without it. I have personally seen it do much good and change lives for the better, both for priesthood holders and for those they serve.

Posts: 5957 | Registered: Oct 2001  |  IP: Logged | Report this post to a Moderator
fugu13
Member
Member # 2859

 - posted      Profile for fugu13   Email fugu13         Edit/Delete Post   Reply With Quote 
It is almost trivial to find extremely high correlations in the social sciences. That has never been a problem. And yet, most of those correlations are not considered particularly useful in an explanatory context, much less a causative one.

If all that is being reported is a variety of correlations, all of which mentioned in the thread have obvious likely other variables driving the correlation, then we gain no information whatsoever. And, as we know the other variables are very likely explanations, the correlations do not increase the evidence of causation at all. Controlling for appropriate variables could easily confound or reverse the correlation, such that there is no causative information in the two-variable correlations.

Posts: 15770 | Registered: Dec 2001  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
quote:
Originally posted by The Rabbit:
Oh, and by the way KOM. If you are talking about the percent of variance in a data set that is described by a linear correlation, you should be reporting R squared and R. Which for graph #4 turns out to be 50.6% not 71%.

I'm not doing a fit, I'm just calculating the correlation coefficient. Since the information content is the same, saying that or the other 'should' be used is purely conventional. If you want to argue that R^2 is more often used in social sciences, that's fine; I wouldn't know. But there's no 'should' about it.

I wasn't comment on which statistic should be used. What I was saying that it that the R^2 is the fraction of the variance in the data set which is described by the correlation the R value is not. Calling the R^2 value a percent makes some sense, it is a percent of the variance captured by the correlation. Reporting the R value as a percent is just wrong. Percent of what?

If your lost post had not been lost, this discussion would have been a lot more pleasant. Perhaps you would like to retract your accusation of dishonesty and/or not reading your posts, now?


Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
Reporting the R value as a percent is just wrong. Percent of what?
Wait, what? You're objecting to the notation? 71% == 0.71. And it's a percentage of correlation, where 100% is "complete linear dependence with no noise". I really do not understand what problem you are having with this. Are you sure we're talking about the same thing? Correlation coefficient, not a goodness-of-fit metric. Possibly they are the same number for the special case of fits to straight lines.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
If your lost post had not been lost, this discussion would have been a lot more pleasant. Perhaps you would like to retract your accusation of dishonesty and/or not reading your posts, now?
And if you had not repeatedly claimed that this paper found that murder rates were correlated with religiosity, you would have deserved more pleasant responses. My insults were directed at a very specific claim you made that homocides were correlated to church attendance. That claim is contradicted by the evidence you used to support it. You deserved to be called on that.

I will retract my accusations that you were either lying or didn't read things carefully when you admit that you retract your claim that this paper provides any evidence to support the your contention that religions increases murder, suicide, teenage pregnancy, STDs and other social problems.

It really is a very very bad example of scientific work and you should feel embarrassed by having used it as a reference.

O, and I still don't believe you picked that graph at random. I've been in science way too long to believe any scientist who claims that they just happened to randomly select the one set of data that best supported their contention. Sorry, I'm not buying it. It would be a bad use of Bayesian reasoning. The chances you picked that graph at random, are 1/45. Based on my past experience, the chance you picked it intentional because you knew it would support your point are closer 3/4. Combining the two, I get a probability of around 0.005 that this was just random chance, so I'm better than 99% confident that this wasn't a random event.

Its just so obvious that this is the graph any one would have picked if they had been trying to make your point. If you retract that exaggeration, I'll apologize for being so snarky about it.

I'm sorry if I'm snarky but I'm in the middle of end of the semester project report grading so my tolerance for this kind of inexcusably bad argumentation is really low. You may be catching wrath that my undergraduate students deserve more readily -- but frankly I expect someone working on a Ph.D in physics to be more compent than that even when posting on an internet forum.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
I already agreed that there is no correlation in homicides. I don't see any way I can convince you I picked at random, so there's no use in further discussion of that; I'll only request that you consider whether I am really stupid or not. Suppose I had decided to pick the graph that best supported my case; that would require me to check out correlations, and I would then be aware that some of them were near zero. If I then reported only the best case, it would be quite trivial to say "Sure, in that one graph; but what about these three"? Which, indeed, is exactly what happened, after half a page of insults. Am I supposed to be completely oblivious to this scenario? Why would I post something knowing I'd have to retract it in a couple of posts?
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
quote:
Reporting the R value as a percent is just wrong. Percent of what?
Wait, what? You're objecting to the notation? 71% == 0.71. And it's a percentage of correlation, where 100% is "complete linear dependence with no noise". I really do not understand what problem you are having with this. Are you sure we're talking about the same thing? Correlation coefficient, not a goodness-of-fit metric. Possibly they are the same number for the special case of fits to straight lines.
Yes, that's what I'm complaining about. Its one of my pet peeves as a scientist. Percent must be a percent of something. It doesn't make sense to talk about percent unless you specify percent of what. R^2 is the percent of statistically meaningful quality of the data. R is not. The fact that 0.71 is 71% of 1 does not mean it is always sensible to call equate 0.71 with 71%. In this case its not. And no I'm not confusing R^2 with a goodness of fit or even statistical significance of the fit.

And yeah, I recognize that I'm probably being overly nitpicky about that for an internet forum. Blame the reports I've been grading. You should still be able to grasp the validity of the point.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
No, actually, I really don't. A correlation coefficient tells you how correlated two variables are. No more, no less. It does not have anything to do with the "statistically meaningful quality of the data", except that the one is apparently the square of the other; which, by the way, ought to tell you that the numbers contain the same information, and which you use is plain convention and not some deep moral quality that matters for whether you are a good scientist. Further, it makes perfect sense to say that two variables are X% correlated, 100% meaning they are the same variable and no noise, 0% meaning one gives you zero information about the other. Perhaps this language is not used in your field; it is in mine, and I really find it quite annoying when you call this meaningless. Differing conventions have no scientific weight. It is not the case that one convention makes you a good scientist and another makes you a bad one.

The percent symbol is just a shorthand for 'hundredths'; whether you call something '0.71', '71 hundredths', or '71%' is completely indifferent. If this is really a pet peeve of yours, you need to rethink it.

I can see, now, where I went wrong; I assumed we were having a reasonable discussion, where one side might say "Take a look at this graph", and the other can say "Yeah, sure, but you've got all these other ones that wash out the significance", without accusations of dishonesty being thrown about; and the first can then again say "Ah, good point; right you are." This is how I generally talk to colleagues when discussing a statistical point. I don't quite know what sort of discussion you were having, but it does not seem to be a polite one.

Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
I already agreed that there is no correlation in homicides.

Yes, you agreed to that after I called you on it. Only a real idiot would have continued to make that argument after I actually calculated the correlation coefficient to be less than 0.1 and I never claimed you were a total idiot. What I want is for you to admit that this paper does not support your original claim that religion causes all kinds of social problems. Its a piece of crap paper and you should feel embarrassed for having referenced it and made unsubstatiatable claims based on it.

quote:
I don't see any way I can convince you I picked at random, so there's no use in further discussion of that; I'll only request that you consider whether I am really stupid or not. Suppose I had decided to pick the graph that best supported my case; that would require me to check out correlations, and I would then be aware that some of them were near zero.

No it wouldn't. It's really easy to tell with a brief glance which of those graphs show significant correlations and which don't. You don't need to do a single calculation to tell that the graphs for child mortality show strong correlations and most of the others do not. To be fare, I don't know that you actually picked the best example. I certainly haven't taken the time to calculate correlation coefficients for all those graphs, but I can easily tell by eye that one there are only 4 or 5 of the 45 graphs that would give you a a correlation as high as the one you picked. So the odds that you would pick one of the best 5 by chance are 1/9. The odds that you would pick one of the because it proved your point and claim it was a random choice, less than 1/4. Which gives me a 2.8% probability that you are exaggerating, so I can only reject your claim at the 95% confidence level not the 99%.

If you happen to be that 1/20 case, I apologize.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
Your priors are made up from thin air; your apology is worth its weight in gold.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
TomDavidson
Member
Member # 124

 - posted      Profile for TomDavidson   Email TomDavidson         Edit/Delete Post   Reply With Quote 
Is there anything more stubborn than a nerd with access to percentages?
Posts: 37419 | Registered: May 1999  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
The percent symbol is just a shorthand for 'hundredths'; whether you call something '0.71', '71 hundredths', or '71%' is completely indifferent. If this is really a pet peeve of yours, you need to rethink it.
I highly recommend you change your opinion on that before you send a scientific paper reporting things as percent or you will find I am not alone in this pet peeve.

If you use percent with out specifying what its a percent of, it can be extremely misleading. If for example you say there was a 50% change in the concentration over time, you have no idea whether that is 50% of the initial concentration or final concentration or the average of the two and it can make an enormous difference. It isn't valid to talk about percent unless you are talking about percent of something meaningful. If a store was advertising a 50% off sale and they'd simply cut the price of everything by $0.50 (50% or a dollar), customers would be hopping mad. Percent is nonsense unless its a percent of something meaningful and that thing is specified. The R value isn't a fraction of some meaningful quantity and so calling a % correlation is just wrong.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by King of Men:
Your priors are made up from thin air; your apology is worth its weight in gold.

If you want to test me on it, I'd be happy to comply.

Here is the proposal.

E-mail me a set of graphs where you have the numbers and I have only a graph. I'll spend less than 5 seconds per graph, and post my estimates of the R^2 values as quickly as I can.

I've done a lot of data analysis. I'm willing to bet you I can estimate an R^2 to one significant digit better than 80% of the time without doing a single calculation.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
If a store was advertising a 50% off sale and they'd simply cut the price of everything by $0.50 (50% or a dollar)
Then they would be using very bad notation. If they'd said $50%, that would be accurate, although customers would still be misled because they would be going against the convention for prices. But it is only a convention, and I would be quite unsurprised to find shops in some other nation using [Our currency symbol]50% without anyone raising an eyebrow. Except that this would be a pretty small discount in any currency I know of, so probably they don't bother.

quote:
The R value isn't a fraction of some meaningful quantity and so calling a % correlation is just wrong.
One more bloody time. 100% correlated variables have a linear relationship with no noise. The statement "X and Y are Z% correlated" conveys numerical information accurately; it is perfectly meaningful.

I'm done with this discussion, it's degenerated basically to semantics. Conventions are fine things, but once it is recognised that the difference under discussion is one of convention, then insisting on the superiority of one over the other is the lowest form of useless pettifoggery.

Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
quote:
If you want to test me on it, I'd be happy to comply.
Not the prior I was referring to. I meant your 25% probability that I was lying. I quite believe you can estimate correlation coefficients by eye, so can I. I also believe you would happily lie about taking only 5 seconds per graph, so your test is useless.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
I also believe you would happily lie about taking only 5 seconds per graph, so your test is useless.
Not if there is a time stamp on the graphs you send me and time stamp on my response.
Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
Puppy
Member
Member # 6721

 - posted      Profile for Puppy   Email Puppy         Edit/Delete Post   Reply With Quote 
I know the subject of a man's position in his home in the LDS church has long passed, but I have to add my two cents, anyway.

Having a "tiebreaker" vote between two people is the same as having the only vote, and I think that approach is ridiculous. In a healthy marriage, a husband and wife should both agree to a decision, or there is no decision.

I see the father's "presiding" role as a ceremonial role. I don't mean that in the colloquial sense of "a role lacking in importance". This is a religion, so ceremony IS important. But that ceremonial role is mostly significant in ceremonial settings — official family meetings, family prayer, father's blessings, that sort of thing. When it comes to making practical choices, there is nothing about being a husband or father that gives a man a right to impose his will on his wife. And a man has no more or less authority than his wife to impose his will on the children. They are equals.

When a man starts thinking that his position in the family grants him some practical ability to get his way all the time, he has completely lost track of what his priesthood is about, and as a result, automatically loses whatever authority he might have presumed himself to have.

Posts: 1539 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Not the prior I was referring to. I meant your 25% probability that I was lying.
You really question this? You are continue to stretch the limits of believability here. I thought it was a fairly generous estimate. It's a prior based on typical behavior not on you in particular. Selecting the best data you ever collected and calling it a "typical result" is so common its become a standard joke. I think its more likely to by 99% of the time than 75%. When humans choose an example to illustrate a point, the tendency to choose an example that illustrates the point well (if that can be readily determined) is incredibly strong. That's one of the reasons for doing studies blinded and truly randomized. If you the reseacher isn't blinded the potential for non-objectivity is extremely high and you clearly weren't blinded. You can't honestly tell me that you didn't know before you interpolated a single number off the graph that the example you picked was going to make the point you wanted. You've already admitted you can tell correlations fairly accurately at a glance.

Maybe you did just happen to open figure 4 first, but do you really expect me to believe that if you had opened figure 3 first you would have bothered to do the work and post it, knowing it would undermine your point? That's what makes this a a non-random choice. I'm don't know whether you looked at all the graphs first and selected one you knew would give a good result or if you opened the first one you saw, recognized it would serve your purposes and then proceeded. Making a judgement about the desirability of the outcome before proceeding makes the selection non-Random. I can't believe you would have proceeded if you had randomly selected a bad example which makes it extremely difficult to accept this was a random example.

I know very well that I would not have bothered reading the numbers off the murder vs church attendance chart if I'd had any question about whether the correlation wasn't going to be pretty close to zero, even though it strengthened my point that this was the example you had used earlier.

Maybe you are more willing to do tedious work to prove you were wrong in an internet forum than I am. But I think the chance of that are much less that 25%.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
Originally posted by The Rabbit:
quote:
Notice that the correlation is still there even if you exclude the outlier US.
No it is not. Did you even look at the figures? Even including the US, there isn't a strong correlation on most of those. From the graphs the only things in that list that are strongly correlated with religiosity are infant mortality and abortion. And once again, correlation does not demonstrate cause. It is much more likely that these things are the result of economic disparity which happens to be correlated with religiosity in the developed countries.
Oh, my mistake. I just found the post I made where I said that infant mortality and abortion showed a strong correlation to religiosity. I made this post an hour and half before you posted the data analysis and well before I posted the insult about your poor reading skills.

So no, I'm not taking anything back. You should be embarrassed for continuing the argument as long as you have.

Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
The Rabbit
Member
Member # 671

 - posted      Profile for The Rabbit   Email The Rabbit         Edit/Delete Post   Reply With Quote 
quote:
One more bloody time. 100% correlated variables have a linear relationship with no noise. The statement "X and Y are Z% correlated" conveys numerical information accurately; it is perfectly meaningful.
KOM, I really don't want to pull rank because I know it won't do any good but this is part of my research specialty, I know my stats extremely well and you are just wrong. R^2 is the percent of variance in Y that can be explained by variance in X. That is statistically meaningful. R is the square root of that. Expressing it as percent is not meaningful because it isn't linearly related to a meaningful statistical parameter.
Posts: 12591 | Registered: Jan 2000  |  IP: Logged | Report this post to a Moderator
King of Men
Member
Member # 6684

 - posted      Profile for King of Men   Email King of Men         Edit/Delete Post   Reply With Quote 
It doesn't matter whether it is "statistically meaningful". If I tell you that the correlation coefficient is 71%, then I have simultaneously told you the value of R^2, unless you are claiming to be so stupid you can't take a square. This is a difference of convention in the presentation! How hard can this be to understand? In physics we communicate correlation coefficients, and we all understand each other. In your field (remind me what it is, by the way?) you communicate R-squared, and all understand each other. Exactly the same information is conveyed. It is just nonsense to say that one is more 'meaningful' than the other. The only time this could possibly be a problem is when we communicate between fields; and even then it's only a problem if one side insists on making it a problem by whining "statistically meaningful" for three pages on end. The square root of a number cannot possibly be less meaningful than the number itself; they are - hah! - 100% correlated.
Posts: 10645 | Registered: Jul 2004  |  IP: Logged | Report this post to a Moderator
  This topic comprises 9 pages: 1  2  3  4  5  6  7  8  9   

Quick Reply
Message:

HTML is not enabled.
UBB Code™ is enabled.
UBB Code™ Images not permitted.
Instant Graemlins
   


Post New Topic  Post A Reply Close Topic   Feature Topic   Move Topic   Delete Topic next oldest topic   next newest topic
 - Printer-friendly view of this topic
Hop To:


Contact Us | Hatrack River Home Page

Copyright © 2008 Hatrack River Enterprises Inc. All rights reserved.
Reproduction in whole or in part without permission is prohibited.


Powered by Infopop Corporation
UBB.classic™ 6.7.2