Not enough Academy applicants?

[snizo]

A clear majority of people in this country no longer support the way the President is handling the war in Iraq. Given that this isn't a popular war ... it doesn't surprise me that the military is having a hard time filling its ranks.

Three figures from non-consecutive years won't tell you if there is a trend of not.

 

[Steve Wilkins]

A clear majority of people in this country no longer support the way the President is handling the war in Iraq.

My glasses must be foggy then, and I don't even wear glasses. Certainly not sure where you come up with a majority, but better yet, how is it a clear majority?

 

[snizo]

A simple google search of "poll support iraq war" returns this link:

Polling Report

(August, 2005) When asked, "Do you approve or disapprove of the way George W. Bush is handling the situation in Iraq?" (CBS Poll, apx 870 people):

57% Disapprove
38% Approve

(August, 2005) Same question, CNN/USA Today/Gallop poll of apx 1,000 people:

59% Disapprove
40% Approve

(Each has a margin of error of +/- 3%)

I would say that 60% is a clear majority. Come on, Steve - this couldn't have come as a surprise. We all know that the country is growing incresingly weary of this war.

 

[Steve Wilkins]

Well of course, of the people who were polled, 60% is a clear majority. The problem lies with using such a small sample size and then making the outcome transfer to the rest of the population. I mean, come on....870 people?....1000 people? Honestly, I thought you were going to use the Newsweek poll as your source that polled 1004 people. I'm tired of hearing people say that the President is living up to their expectation or that they disapprove of how he's handling the war, and yet when asked, can never deliver a coherent strategy of execution of their own. I'm not saying this is you....maybe it is, I don't know. I will have to dig up a good article I had to read for one of my classes that dealt with polling issues.....who are polled, numbers, wording of questions, etc.

 

[squeeze]

Well of course, of the people who were polled, 60% is a clear majority. The problem lies with using such a small sample size and then making the outcome transfer to the rest of the population. I mean, come on....870 people?....1000 people? Honestly, I thought you were going to use the Newsweek poll as your source that polled 1004 people. I'm tired of hearing people say that the President is living up to their expectation or that they disapprove of how he's handling the war, and yet when asked, can never deliver a coherent strategy of execution of their own. I'm not saying this is you....maybe it is, I don't know. I will have to dig up a good article I had to read for one of my classes that dealt with polling issues.....who are polled, numbers, wording of questions, etc.

pretty common knowledge, but since you'll ask for a source anyhow.. here:


"As the sample size increases there are diminishing returns in percentage error. At percentages near 50% the statistical error drops from 7 to 5% as the sample size is increased from 250 to 500. But if the sample size is increased from 750 to 1,000 the statistical error drops from 4 to 3%. As the sample size rises above 1,000, the decrease in marginal returns is even more noticeable."

source: http://www.ropercenter.uconn.edu/pom/polling101.html#scient
or
http://www.robertniles.com/stats/sample.shtml
or
http://www.surveysystem.com/sscalc.htm
or
http://www.busreslab.com/tips/tip36.htm

 

[DanMa1156]

I think we should also be aware that the polls are referring to the way that the President is handling the war in Iraq, as opposed to being for/against the war itself.

 

[Steve Wilkins]

It's also pretty common knowledge that you can make statistics do whatever you want them to do and to prove whatever you want to prove.

I think we should also be aware that the polls are referring to the way that the President is handling the war in Iraq, as opposed to being for/against the war itself.

Yes, I understand what that particular poll question is asking.

 

[wink]

It isn't the sample size that matters (as long as the margin of error is supplied), because the sample size largely effects only the margin of error. 800-1000 is a perfectly acceptable sample size. What matters is the construction of the question, obviously, and the demographic of the sample. Sample demographics is where some of these polls fall apart.

 

[IRfly]

In statistics, if you have a truly random sampling, then you'll get approximately the same results from a sample population of 30 as you will from 3000. Of course, a truly random sampling is impossible, but by the time you're over a sample population of 120 or so, there is going to be no appreciable difference in the results of the poll, no matter how much higher the sample population is. The guys who do this are pretty darn good at their jobs.

 

[wink]

In statistics, if you have a truly random sampling, then you'll get approximately the same results from a sample population of 30 as you will from 3000. Of course, a truly random sampling is impossible, but by the time you're over a sample population of 120 or so, there is going to be no appreciable difference in the results of the poll, no matter how much higher the sample population is. The guys who do this are pretty darn good at their jobs.

There can be appreciable differences in poll results despite a "random" sample. How do you define random? Is the random sample taken from people that answer the phone at one o'clock in the afternoon when most people with jobs are working? Is it weighted in any way? Most polls are. Weighted for gender, political party, geographic part of the country etc. How about people that have gone wireless or screen all their calls or are too poor to have a phone? They are part of an important demographic that goes un polled. All that said, I agree, most pollsters do a good job because if they get it wrong they won't get rehired. The is one large caveat though. Some polls are no different then the testimony of an expert witness in a trial. You can find an expert to testify to almost anything you want to support in court. Pollsters do the same thing. If for instance they know a poll is being commissioned by an organization with an agenda and that organization wants the poll to publicly support their agenda, the pollster will give them something they can publish. If they don't, the money spent on the poll is wasted and the pollster will not be rehired. The pollster does not have to lie to make their client happy. They just manipulate the sample and questions to provide the result they want. That is why during elections, internal polling is more accurate (and sometimes kept secret for that reason). The politcal parties truely need the most accurate data posssible for strategic reasons. External polls by the same party may give a different result then internal ones.

 

[The Chief]

Mea culpa. You guys are tough. But correct.

I said random sample, Of course it is not a random sample, in fact, not such thing as a random sample.

In fact not even a sample at all, it is in fact a few data points. Data are accurate (have source).

Let me add one more data point, if I may:

Class lf 1998:

Applicants 11,340
With nominations 5,051
Qualified Nominees 1,959
Offers of Admission 1,555
Admitted 1,207

I guess the point is there is indeed no trend, but in fact, the number of applicants vary from year to year.

 

[IRfly]

There is so such thing as a truly random sample. The only point that I wanted to make was that if a poll is somehow skewed, then it doesn't matter how many you poll...After 120 or so, the bias will be consistent no matter how large your pool gets. As far as your other comments, Wink, I think that probably the most important is that of how the question is phrased. If you see certain names attached to a poll (Gallup, for instance), you can be fairly assured of the poll's fairness and accuracy. The question, though, can be crucial. I think that "Do you approve or disapprove of George W. Bush's handling of the war in Iraq?", however, is pretty straightforward. It's also a very narrow-scope question. To disapprove of Iraq doesn't mean that they wouldn't vote for him again, for instance, or that they approve of anyone else's plan more.

 

[snizo]

I'm tired of hearing people say that the President is living up to their expectation or that they disapprove of how he's handling the war, and yet when asked, can never deliver a coherent strategy of execution of their own.

You've got to be kidding me. So no one is allowed to say 'this is broken' unless they can say 'this is how you fix it'? That is rediculous. I'm sure there are plenty of people whose 'strarategy' is quite simple - they would have never gone to Iraq in the first place.


(On a seperate note) Whoever said statistics can be made to say whatever you want is somewhat correct. But the reason is the same that jouranlists can use selective quoting to spin what you say any way he wants to. In this case, we were given the exact wording of the question. It could be biased if the pollster told the person who answered a story about a document he just learned about that the President had that proved the weapons of mass destruction didn't exist. Most of us assume that didn't happen, though, because as IRFly said - these pollsters are experts with statistics, too, and are very good at their job. If their opposition can find a reason to invalidate the pollster's findings, they will - and the pollster is out of a job permanently in this business.

On a side note - random samples exist, but you will never see a poll based on one. Even if you were to use a computer to randomly pick number to form a phone number (and computers are incapable of forming a truly random number) and call - you still can get a biased conclusion. If you call only between 6 and 8pm, then you are probably only getting people who work normal white collar jobs with 9-5 hours. If you use the phone, you are excluding people who do not have phones or who do not answer calls from people they don't know (caller ID). Etc etc etc. We just do the best we can with polling and poll those who are most likely to know and care about the topic enough to have an answer to the question.

 

[wink]

To disapprove of Iraq doesn't mean that they wouldn't vote for him again, for instance, or that they approve of anyone else's plan more.

Or that you didn't approve of the war in the first place, or that you still approve of the war, just not how the Pres is handling it, or that you would or would not approve of immediate withdrawal, etc. May be a straight forward question, just doesn't tell the whole story.

Wow, what a threadjack.

 

[wink]

You've got to be kidding me. So no one is allowed to say 'this is broken' unless they can say 'this is how you fix it'?

To continue with your analogy, if I set the braks on my bright silver jet and tell the mechanic it is broke I don't have to tell him how to fix it. That is a true statement. But if I am to look over his shoulder and tell him and every other mechanic in the shop that he is doing it all wrong, that I don't approve of his handling of the problem, I damn well better be able to pull out my A&P liciense and flash the resume that proves I worked on transport catagory aircraft before. Otherwise, my opinion means nothing. Ya, everyone is entitled to an opinion. Most just aren't informed opinions.

If their opposition can find a reason to invalidate the pollster's findings, they will - and the pollster is out of a job permanently in this business.

Not true. I have seen expert witnesses torn apart on the stand and they keep getting jobs. All it takes is a couple jurors to believe their junk science/engineering opinion and the attorney who hired him gets his moneys worth. Same with bought off boutique pollsters. It doesn't matter if the opposition can tear the poll apart. It gives the pollster's client talking points and bullets to use on every cable show that will have him or in every paper that will quote him without the opositions revelations to counter it.