Gallup poll debunks myths about American Muslims

[NotLostJustWandering]

http://www.salon.com/news/politics/war_room/2011/08/06/gallup_muslim_americans

Only those whose views of Islam derive from prevailing stereotypes should be surprised to learn that the overwhelming number of Muslim respondents said that the killing of civilians in war is never justifiable. After all, the Qur'an expressly forbids it. What shocks me is that the majority view of EVERY OTHER RELIGIOUS GROUP POLLED said that such killing is sometimes justifiable. Only among the atheists and agnostics was the majority view on this moral issue the same as the Muslims'... but not by nearly as wide a margin.

http://www.salon.com/news/politics/war_room/2011/08/06/gallup_muslim_americans/military_civilian_deaths_.jpg

[Pasadenacpl2]

The INTENTIONAL killing of civilians is never justifiable. Things like IEDs, and homicide bombings are never justified. But, sometimes, no matter how hard we try, non-combatants die. It is tragic, but an inevitability of war. I am thankful that we go out of our way to not kill civilians. No army has ever invested so much in money and loss of manpower to ensure the fewest number of civilian casualties as ours does.

Would that Al Qaida did the same.

Pasa

[Hephaestion]

In statistics there are 2 types of error (type 1 and type 2). One of them says accept the hypothesis under test when the evidence is in fact wrong the other is to reject the hypothesis when the evidence is right. Polls are subsets of a population's response. Was the subset truly representative in a land where there is so much diversity? Populations are fickle in composition and outlook. Populations may be pursaded to give an answer adapted from a plausible question claimed later to be related. There are lies damned lies and statistics.

Islam can be as aggressive as any other religion that seeks to pursuade the infidel of the error of their ways.

.

[Annika L]

In statistics there are 2 types of error (type 1 and type 2). One of them says accept the hypothesis under test when the evidence is in fact wrong the other is to reject the hypothesis when the evidence is right. Polls are subsets of a population's response. Was the subset truly representative in a land where there is so much diversity? Populations are fickle in composition and outlook. Populations may be pursaded to give an answer adapted from a plausible question claimed later to be related. There are lies damned lies and statistics.

Islam can be as aggressive as any other religion that seeks to pursuade the infidel of the error of their ways.

.

In every instance, the likelihood of those errors is reduced by using high sample sizes. Generally, a sample size about 30 is sufficient to be meaningful. 50-100 is seen as a pretty darned large sample.

To quote the technical report for this study:



The data from the Gallup-Healthways Well-Being Index are reported
out using aggregated data, as well as most recent data. For purposes
of this study, aggregate data refer to interviews conducted from
January 1, 2008-April 9, 2011. The total sample based on aggregate data
is 868,264 adults, of which 3,883 self-identified as Muslim Americans.
In the study, most recent data are based on interviews conducted from
January 1, 2010-April 9, 2011. The total sample based on these data is
336,888 adults, of which 1,492 self-identified as Muslim Americans.
Data from the Muslim American population were collected from
February 10, 2010-March 11, 2010, and October 1-21, 2010. The total
sample based on the combined Muslim-American polls is 2,482 adults,
of which 475 self-identified as Muslim Americans.

Yes, there are always issues with statistics. How questions were asked, how samples were selected, reporting bias, etc. are always valid concerns.

But I'm not worried about sampling error here. Nor is a poll a hypothesis test.

[Gearbox]

When a question that leads to numerable possible scenarios is asked, the interpretation of the Yes/No/Maybe answers can also have numerable possibilities.

I don't think that anybody would expect soldiers to not retaliate against armed aggressive civilians for eg.
With that in mind, the poll results may be interpreted as 'how much American Muslims want the American military to get shot etc.'.:rolleyes:

[Pasadenacpl2]

I don't have a problem with the study itself. Though on the particular question being discussed here, they say right in the article that they asked the question in different ways with several different wordings. This boggles my mind.

Pasa

[tenni]

"Though on the particular question being discussed here, they say right in the article that they asked the question in different ways with several different wordings. This boggles my mind."

This is a fairly standard practice. It is a form of double blind testing/surveying. I believe that it is often used in attitudinal surveys and psychological evaluations.

[Hephaestion]

When one uses a poll as indicative of the population at large it is in fact a statistical sudy. In other words the sample is declared 'representative' and must therefore have confidence limits.

To suggest that a mere 100 people constitutes a large sample size and can represent a nation of several million is a little presumptuous.

To quote sample sizes heading towards the million mark is still presumptuous that these still represent what a wider number think. The claim will be that the samples are chosen from representative areas so as to be representative is circular and prejudging the outcome.

Reminder, it's type 1 and type 2 errors, diversity, fickleness, and loaded questions. Sometimes it is even corruption. Wasn't GWBush's election marred by the suspicion

There have been many polls on here some claiming to be academic studies where the real answer is a category which is e.g. not given. Same old problem classifiying continuums into simplistic little answers, better known as stereotyping and propaganda.

However, the real bugbears are the resurrection of bum coefficients such as IQ. Radio 4 was fooled with this one when it was declared by a bogus market research firm that the intelligence of Internet Explorer users was 'below average' as measured by IQ; as if one can measure intelligence or even morality.

[Annika L]

*sigh* Heph, I generally have such respect for you and your postings, but here I don't think you know what you're talking about.


When one uses a poll as indicative of the population at large it is in fact a statistical sudy. In other words the sample is declared 'representative' and must therefore have confidence limits.

I didn't say it wasn't a statistical study. I said it was not a hypothesis test. One could perform a hypothesis test based on these data, but no such test was discussed, and no such results or conclusions were drawn.

Regarding confidence limits, one could construct a confidence interval for each of the groups' responses, but given the huge sample size, it is doubtful that such intervals would shed much light beyond the charts shown, since the intervals would be rather narrow. Yes, it would have been nice if they'd given the margins of error anyway...not sure why they didn't (if they didn't)...maybe it's buried somewhere in the report.


To suggest that a mere 100 people constitutes a large sample size and can represent a nation of several million is a little presumptuous.

This statement indicates ignorance (or perhaps rejection?) of probability and statistical methodology. For most purposes, a random sample of 100 individuals is indeed quite large. When the sample size reaches several hundreds or even thousands, as in this study, it's definitely large enough to draw meaningful conclusions about questions with three possible answers.

As I said before, all statistical studies have their limitations and areas of concern. If you care to go into the report and point out some genuine concerns about how things were done, feel free to do so.

But learn something about probability and statistical methodology before you try to attack it, rather than just trotting out trite lines and quotes about statistics.

[tenni]

The data included in this report come from three independent sources based on self-identified religious affiliation: The Gallup Healthways Well-Being Index (also referred to as the Gallup Nightly or G1K), as well as two independent studies of the Muslim-American population.


The Gallup-Healthways Well-Being Index was developed to obtain statistics on the state of wellbeing for adults in the United States. The endeavor accumulates responses from a random minimum sample of 1,000 U.S. residents, 350 days per year. On any given evening, approximately 250 Gallup interviewers conduct computer-assisted telephone interviews with randomly sampled respondents 18 years of age and older, including cell phone users and Spanish-speaking respondents from all 50 states and the District of Columbia. The survey includes many of the standard demographics, including race, income, education, employment status, occupation, and household size (number of adults). Location data, such as zip codes, allow researchers to map the responses to particular parts of the country and accumulatedata for local-level comparison and interpretation...........more than 90% of the U.S. adult population is represented in the sample..............The total sample based on these data is 336,888 adults, of which 1,492 self-identified as Muslim Americans.

[Hephaestion]

The hypothesis in a poll is that the results are representative. As pointed out this can be flawed assumption for a variety of reasons.

Then there is the mistake that can be made as pointed out at the outset Type 1 and Type 2 errors. Surely this is common sense but is formalised in this categorisation.

100 is a woefully small sample to allow generalisation about a population of several millions what ever the school text books say.

Question yourself as to why it is that I can say these things with the utmost confidence.

[tenni]

Heph
The sample size was 336 888 and not 100.

Why not go and examine the methodology section of the actual study? You can find it by going to the article and then click the red report (pdf) button that will give you the actual report.

[Hephaestion]

I know that Tenni.

The point of debate here is whether 100 is a large sample size. It can be in certain circumstances such as where a particular mathematical model is being used as the standard e.g. in parametric tests and the investigaiton is limited in its scope. There are times when it is woefully inadequate.

This started out with, of the religious groups only muslims felt that killing civilians was unjustifiable. read between the lines. The poll was asking muslims living in the USA who must feel that they are treading on eggshells whether they believed in the killing of innocent civilians as being justifiable.

Were they to say yes then would they not have identified with such acts 911 i.e. the killing of innocent civilians. Suspected bias straight away.

[Annika L]

The hypothesis in a poll is that the results are representative. As pointed out this can be flawed assumption for a variety of reasons.

Then there is the mistake that can be made as pointed out at the outset Type 1 and Type 2 errors. Surely this is common sense but is formalised in this categorisation.

100 is a woefully small sample to allow generalisation about a population of several millions what ever the school text books say.

Question yourself as to why it is that I can say these things with the utmost confidence.

No, Heph, there is no hypothesis being tested in a poll. To be meaningful, the sample the poll is given to must be able to be considered representative, but this is far from a hypothesis, and is not tested in any case. Rather, people who read the study must decide for themselves whether the sampling methodology and sample size are sufficient to make the results meaningful. And because there is no hypothesis test, the notion of Type I and II errors simply does not apply.

I'm afraid the only reason I've been able to come up with why you can say these things with utmost confidence is that you must be ignorant of the mathematical principles involved. It's all I'm left with, and each statement you're making here adds to this impression. Read a bit about probability and sampling distributions. Or start talking sense.

If we were talking about a question where there was a large diversity of possible (and discrete) answers, I'd agree that 100 is too small a sample to capture the diversity of opinion from a large population. But as the question was phrased, there are three possible opinions (Never, Sometimes, and Depends). Under those circumstances, for most studies, a sample of size 100should be fine.

But as tenni points out and as I've pointed out, this particular study uses a sample size *far* larger, and gives me no reason for concern as to sampling error (in fact I did an off-the-cuff estimate of the margin of error, and came up with less than half a percentage point with 95% confidence).

That is very different, however, from saying that there is no *response bias*, and I've already said a time or two that this is a valid concern (for precisely the reason you point out to tenni). But increasing the sample size would in no way remove that response bias...it's just there, and yes, it does alter what light you read this in to draw conclusions.

But response bias is unrelated to reprentativeness of the sample and significance of results.

[Hephaestion]

Anika-L ".....That is very different, however, from saying that there is no *response bias*, and I've already said a time or two that this is a valid concern (for precisely the reason you point out to tenni). But increasing the sample size would in no way remove that response bias...it's just there, and yes, it does alter what light you read this in to draw conclusions.

But response bias is unrelated to reprentativeness of the sample and significance of results....."

The survey is nonsense as is the last sentence above.
I am floored by the persistent notion that a sample size of 100 individuals can be seen as representative of an nation of several millions especially when one is dealing with opinions.

Confidence limits of 95% - based on Monte-Carlo methods? Should one assume that you have access to the original responses? Do any of the multivariate methods deployed rely on parametrics? Were matching coefficients used if so which ones and were they profiled?

Even simple proportions can be statistically viewed e.g. Observed versus Expected (Expected that there will be no differences, Observed claimed that there are and could these differences have occurred by chance - the larger the sample size then the better the likelyhood that the results are real and representative) at that moment.

Incidentally there is also a temporal shift in the acquisition of data and the assumption that there will be no influence in an age of rapid communication. A couple of nights ago there were no riots in the UK. Currently there is a jaded view of a common section of our communities with an interest in many of the remainder in deadly response.

Reminder that NLJW's claim from the survey was that 'of the religious groups only (ONLY) muslims felt that killing civilians was unjustifiable'. Does Jedi count as a religious group - was that polled? We are dealing with reality here.

It doesn't matter the niceties or otherwise of the statistical results.
The very foundation of the survey is suspect i.e. the responses are unreliable.

Type 1 and type 2 errors are always things to bear in mind
One of them says accept the hypothesis under test when the evidence is in fact wrong the other is to reject the hypothesis when the evidence is right.
The concept itself is didactic.

[NotLostJustWandering]

I know that Tenni.

The point of debate here is whether 100 is a large sample size.

A debate between you and a straw man, completely irrelevant to the topic of this thread.



This started out with, of the religious groups only muslims felt that killing civilians was unjustifiable. read between the lines. The poll was asking muslims living in the USA who must feel that they are treading on eggshells whether they believed in the killing of innocent civilians as being justifiable.

Perhaps, but, as I said in my first post, the Qur'an expressly forbids the killing of civilians in war. Are you contending that Muslims' opinions are not influenced by the Qur'an?


Reminder that NLJW's claim from the survey was that 'of the religious groups only (ONLY) muslims felt that killing civilians was unjustifiable'. Does Jedi count as a religious group - was that polled? We are dealing with reality here.

I specifically said "of the religious groups POLLED." And since, as you say, we are dealing with reality here, no, there is no such religion as "Jedi."

[NotLostJustWandering]

The INTENTIONAL killing of civilians is never justifiable.

No one goes to war unintentionally.



Things like IEDs, and homicide bombings are never justified. But, sometimes, no matter how hard we try, non-combatants die. It is tragic, but an inevitability of war. I am thankful that we go out of our way to not kill civilians.

And "our way" is necessarily attacking other nations halfway around the world?


No army has ever invested so much in money and loss of manpower to ensure the fewest number of civilian casualties as ours does.

Lies and propaganda. These days the top priority of the U.S. Army is to minimize loss of its own soldiers' lives (and to distract attention from the deaths that do occur) in order to prevent large-scale anti-war protests such as happened during the Vietnam War. Hence the preference for high-tech air attacks to man-to-man ground combat -- and the subsequent high odds of civilian deaths.



Would that Al Qaida did the same.

Are you saying Al-Qaeda exemplifies Islamic values or represents the views of American Muslims? If not, why do you think this point germane to the discussion at hand?

[sammie19]

I specifically said "of the religious groups POLLED." And since, as you say, we are dealing with reality here, no, there is no such religion as "Jedi."

In the last but one census held in this country many thousands of people gave as their religion "Jedi". Were they telling porky pies?;)

[NotLostJustWandering]

In the last but one census held in this country many thousands of people gave as their religion "Jedi". Were they telling porky pies?;)

Well, I'd say they were taking the idea of religion less seriously than the average American does. I'm sure there are a few Americans who might identify their religion as such, and they are probably very serious about it (the Church of Elvis people ain't joking, either). But they wouldn't amount to a statistically significant number, which makes Heph's complaint very suspicious...

[Hephaestion]

I am a Jedi and believe in the force

Bzzzzzt

I shall bite my tongue when it comes to skyscraoers and aeroplanes and innocent civilians and muslims.

[tenni]

"A debate between you and a straw man, completely irrelevant to the topic of this thread."

What I find interesting about some posters is how strongly they prefer to get into a debate about the validity of polls and scientific research rather than what the poll/survey/research was actually stating.

In the article or actual report, I believe that it stated that US Muslims could identify with the victims of wars in Iraq and Afghanistan than other US people. They gave that as a reason/hypothesis as to why US Muslims did not support the killing of citizens more than others in the US.

[Annika L]

Anika-L ".....That is very different, however, from saying that there is no *response bias*, and I've already said a time or two that this is a valid concern (for precisely the reason you point out to tenni). But increasing the sample size would in no way remove that response bias...it's just there, and yes, it does alter what light you read this in to draw conclusions.

But response bias is unrelated to reprentativeness of the sample and significance of results....."

The survey is nonsense as is the last sentence above.
I am floored by the persistent notion that a sample size of 100 individuals can be seen as representative of an nation of several millions especially when one is dealing with opinions.

Confidence limits of 95% - based on Monte-Carlo methods? Should one assume that you have access to the original responses? Do any of the multivariate methods deployed rely on parametrics? Were matching coefficients used if so which ones and were they profiled?

Even simple proportions can be statistically viewed e.g. Observed versus Expected (Expected that there will be no differences, Observed claimed that there are and could these differences have occurred by chance - the larger the sample size then the better the likelyhood that the results are real and representative) at that moment.

Incidentally there is also a temporal shift in the acquisition of data and the assumption that there will be no influence in an age of rapid communication. A couple of nights ago there were no riots in the UK. Currently there is a jaded view of a common section of our communities with an interest in many of the remainder in deadly response.

Reminder that NLJW's claim from the survey was that 'of the religious groups only (ONLY) muslims felt that killing civilians was unjustifiable'. Does Jedi count as a religious group - was that polled? We are dealing with reality here.

It doesn't matter the niceties or otherwise of the statistical results.
The very foundation of the survey is suspect i.e. the responses are unreliable.

Type 1 and type 2 errors are always things to bear in mind
One of them says accept the hypothesis under test when the evidence is in fact wrong the other is to reject the hypothesis when the evidence is right.
The concept itself is didactic.

Ok, I am now convinced that you have no idea what "response bias", "representative sample", or "statistical significance" mean, let alone the difference between "confidence limits" and a "confidence level".

Are we supposed to be impressed that you've heard terms like "Monte-Carlo methods", "parametrics", "matching coefficients" or "multivariate methods"? When you misuse them or pull them out of thin air, it does nothing to help you. Given your attitude and lack of knowledge here, I'm done with this part of the discussion...one cannot argue sensibly with the uninformed, unless one is prepared to inform them...and I haven't the time or patience to give you a free course in statistics (and you don't want one).

Basically your message on this thread seems to reduce to: I don't like the implications that are depicted in this poll, there are reasons to question the results, and I don't know enough about statistics to trust it as a tool for showing anything.

Of those three bits, the one with some validity is that there are reasons to question the results...because of response bias. I've never said anything in even vague disagreement with that. But I think that despite the response bias, the results (particularly for Muslims and for Atheists/Agnostics) are at least interesting, and worthy of consideration...far from utter nonsense. If you disagree, that's fine. But don't throw around statistical terms to try to enhance the image of your opinions. Back up your opinions with real argument, and if you're going to attack statistics, know the methodological and/or mathematical issues you're attacking.

[Hephaestion]

OK Annika-L you think I talk bollocks. I think the same of you.

At one time you tell me that there are no statistics involved and then you claim that there is a 95% confidence limit involved in the off the cuff assessment that you have made of the result. I am amused.

Please do not think that I am trying to impress anyone. I wouldn't waste the effort. I merely draw attention to the supercilious claims of rubbish polls such as the one presented. They have curiositiy value but that's it.

I have given a multitude of reasons to doubt the claim that muslims in America are any more moral than people of other faiths in the same country. However, you just refuse to see it.

.

[Annika L]

OK Annika-L you think I talk bollocks. I think the same of you.

At one time you tell me that there are no statistics involved and then you claim that there is a 95% confidence limit involved in the off the cuff assessment that you have made of the result. I am amused.

Please do not think that I am trying to impress anyone. I wouldn't waste the effort. I merely draw attention to the supercilious claims of rubbish polls such as the one presented. They have curiositiy value but that's it.

I have given a multitude of reasons to doubt the claim that muslims in America are any more moral than people of other faiths in the same country. However, you just refuse to see it.

.

LOL, you silly dear. As I pointed out earlier, I never said there were no statistics involved; I said and maintain that they cited no hypothesis test, so there can be no Type I or Type II error.

But on what grounds do you claim that I don't see that there are reasons to doubt that American Muslims are more moral than people of other faiths here? I've repeatedly agreed that response bias is an issue here, and acknowledged that there are concerns with any statistical study.

Be amused by all means, but learn to read.

[tenni]

"I have given a multitude of reasons to doubt the claim that muslims in America are any more moral than people of other faiths in the same country. However, you just refuse to see it."

The survey never asked people nor stated which group was more moral than another group. The report made no moral judgment but reported beliefs of people from various self reported religious beliefs. The above statement is an extrapolation and interpretation of the survey.

The survey found out is that the largest self reported group ( 78 % of US Muslims) believed that it is never justifiable for the military to target and kill civilians.

[Hephaestion]

That's right Annika-L you get abusive and drum your feet

Statistics do not and never have proved anything. What one does with statistics is to look at the result and then ascertain by whatever method is appropriate whether the result is so extreme that one would be wise to take note of it

Even if results are extreme and indicate that the picture is real it should be noted that one is still at liberty to accept or reject the findings. Obviously some justification needs to be presented to support the decision

Type 1 and Type 2 errors are highly pertinent in concept which is what I said; one accepts the result when it is wrong, the other rejects the result when it is right.

All of the religions in our concepts are fundamentally moral and share the same aspirations: be nice, don't kill, don't steal and above all don't bugger you neighbours' oxen.

Here, the starting assumption in the survey must have been that there was no difference in the proportions of people deciding to kill civilians. The title given is "Gallup poll debunks myths about American Muslims" So when NLJW trumpets that American muslims are more moral than other religious groups on this matter one needs to look at the result carefully.

The underlying assumption for such a claim must be that the survey was representative of the entire nation (there is also the knock on effect on other nations and their perceptions).

In finding representative areas one needs to profile the nation in some fashion. This is where the multivariate analyses come in. Underlying this assessment is whether there are metrics or characteristics which are of use in doing so. It is possible that the factors involved mislead. If one has already done the demogrpahic analysis then is it up to date? This is a serious issue. When one receives the results these may well be reanalysed in term of superimpositon above the demographic distributions. Is there congruence is the basic question. Is(are) there some other correlation(s). Again a multivariate approach may be used.

The number of people involved in the survey was in the hundreds of thousands. Laudable! So when the suggestion is made that a sample size of 100 is adequate hot tea is snorted from nostrils.

The obvious thing to look at is what was asked and how it was done. THe categories present were: Muslims; Protestants; Catholic; Jewish, Mormons, Atheist/Agnostic. Screw the rest e.g. native Americans, Orthodox Christians; Bhudists, Orientals, and 'modern day' religions e.g. Jedi. There seems to be selective inclusion/rejection here.

In order to sample such a large number of people one assumes that temporal shifts were held to a minimum. However, the survey must have taken some time.

The concept of loving thy neighbour has taken a serious dent in the UK with riots on the face of it initiated and continued by unruly negros. "It's the fucking nig nogs again" is what one hears despite the evidence that this is not the only group involved. That took just 2 days and the debate about armed police pales into insignificance (although one doubts whether the familiy of Mr Menendez have forgotten). One receives news from around the world that the moral policemen of the world are being killed by these ungrateful savages in cowardly civilian attacks. One side says that this is an army yet they do not wear unforms and are prepared to sacrifice themselves and their countrymen. Propoganda? It will certainly colour the thought processes of participants in the survey, if there is an interest in participation.

When surveys are carried out there are two broad methods. One may be immediate (Telephone, or face to face). The other may be by submission e.g. via letter or Internet. There ill almost certainly be a series of filtering questions, awkward category fittign, discardng of troublesome answers; we've seen it here.

Someone contacts a person and asks 'are you prepared to kill civilians' and by the way we are the reputable Gallup organisation. One imagines that Gallup are not an organisation that asks these questions without sponsorship. Who or what is in the background? Why are you asking must feature high in people's thought processes. Empthy or antipathy (Tenni). Self identification of being muslim? Arab in lineage, Pharsee, Negro, refugee born abroad, born in the USA?

Is this an attempt to deflect the concept of danger and thus allow the US government to withdraw its forces 'wth honour'? (heard that one before?).

Had the poll returned that there was no difference in the groups given then there may well have been lack of scrutiny. But 'no news, does not sell'

So many questions on the vlaidity of the claim

[sammie19]

I think it likely that when we are in a society which has its suspicions and neuroses about us and what we believe, it is a natural defensive mechanism to answer in terms which may not necessarily be honest. Self preservation is a powerful emotion, and I would be fascinated to know how moslems would have answered the poll if the war on terror was on Buddhists, Jews, or Hindus.

Equally I wonder what the responses would have been from Christian and Jewish groups if it was they who were in the place of Islamic people?

I don't give any credibility to this poll because I don't think like most polls, it has extracted honest reactions from far too many people.