'Major poll shows gun ownership in decline'

[Glenn E. Meyer]

I won't claim to understand the arcana of statistical analysis, but I have always had my doubts about how any poll (sample) of a thousand or two thousand people can actually be accurately representative of the over 300 MILLION people in this country.

The idea was that the standard error of proportion is based on a division by the square root of the sample size. When you got above 1200 participants, increasing sample size didn't really increase power that much usefully. One problem with megadata is that everything is now statistically significant due to the increase in sample size and shrinking of the standard error. Other methods are being devised and NHST is seen on the way out. NHST = null hypothesis significance testing. So all those hated stat courses are obsolete. Hahaha!

Taught that crap for years!

Anyway, if you buy into the 1200 subject argument, the validity of the sample becomes paramount. It assumes a sample that mirrors the population. With a 300 million size group - as pointed out a representative sample of a resistant sample is quite a problem.

Gun research is fraught with biases and you have to know the honesty of the players.

[SIGSHR]

IMHO people answer polls differently face to face than anonymously.

[ballardw]

Yes you will get somewhat different answers depending on face-to-face (FtF), telephone or paper/web surveys.

Possible causes:
FtF and phone may have the interviewer clarify questions.
FtF and phone may have bias introduced by those clarifications.
FtF and phone interviewers may attempt to get clarification or try to overcome initial refusals or try to keep respondent cooperating.
Verbal emphasis in the way Ftf and phone surveys are conducted may change impact on respondent.
Paper / web surveys rely more on respondents reading level.

We won't even go into the bits where interviewers make data entry errors accidentally or intentionally.

And lots more differences.

[mag1911]

One of the classic statistics assignments is determining the percentage of white vs black marbles in a bathtub. The problem assumes even distribution of the marbles and the sample taken from all areas inside the bathtub. The percentages can be mostly determined with a relatively small sample size vs the total marble count in the bathtub.

But people aren't marbles and the US isn't a bathtub. Getting a truly even distribution of responders and truly unbiased responses is almost impossible even if pollsters make the attempt but most don't have the time/money for that.

Most pollsters are hired and may bias the poll toward the customer's expected outcome. Most responders tend to nuance their answers to controversial subjects. That's when you see polls showing maybe 70% for/against but the actual outcome much different.

[tirod]

It's an intensively focused survey of non gun owners in a anti gun Metro conducted as a teaching tool by students learning statistical techniques. There is no basis for real world application of the results.

Nothing in this survey is worth discussing, it's completely junk. Better to discuss which one of Bloombergs shadow groups handed the cash to NORC to conduct it and get the obvious answers they would.

It's not about statistical variance and probability. They knew going out there would be extremely slanted views they would gather, as very few would admit in public they owned a firearm in a jurisdiction that is anti gun.

Same survey in Manhattan would get even worse percentages.

"Hi, we're interviewing the man on the street, do you own guns?" "Why yes, I do, in fact I'm carrying a new AR Pistol I just finished assembling right here in my tennis bag, even has the Brace on it. "

Think about it - concealed is concealed, in Metro Chicago IIRC there's only one or two gun stores at all.

Of course ownership is "declining." But that is about to change dramatically.

[skoro]

would YOU tell a stranger on the phone you own guns?

I'd have to answer no, because I don't have any guns.

I only own firearms.

[mag1911]

I believe this an example of "the customer is always right" method of polling:

http://www.kansas.com/opinion/opn-co...e15022460.html

Getting 98% of a truly "random" sample to agree on anything is quite a stretch. I also don't see this question in the poll. A couple links in the editorial show the polling method.

[Glenn E. Meyer]

Interestingly on polls, the Israeli election just went against the polls as those polled probably refused to answer.

Just an example of polling, not interested in that outcome here.

[BWM]

All I go to say B H O has been the best gun sale man in years.

[Andy Blozinski]

Polls can be fun to read. You really have to get multiple data sets. Oftentimes it's also useful to get polls about related subjects to support a pattern.

Here's the result of a poll on a related subject:
http://news.yahoo.com/gun-control-ad...202659404.html

The last paragraph is the relevant connection:

"By a margin of 52 percent to 46 percent, Americans say protecting the rights of gun owners is more important than gun control, according to a survey by the Pew Research Center released in December. The survey, which came just two years after the massacre at Sandy Hook Elementary School in Newtown, Conn., was the first time Pew found more support for gun ownership than gun control in more than two decades of surveys on the issue."

[barnbwt]

"Interestingly on polls, the Israeli election just went against the polls as those polled probably refused to answer."
It seems there's a bit of evidence piling up that there was some shady doin's going on as well with regards to us "swaying" the voice of the electorate over there proactively against the incumbent, which may (did) have had an effect on how poll's shook out in the days ahead of the election.

As they say, "the democratically elected government of so-and-so won't overthrow itself"

TCB

[Andy Blozinski]

Polls can change pretty damned fast too. Sometimes people will answer differently when it comes to actual crunch time.