For Ed Huntress - file request

Ed,
I tried to send this to huntres23, but the mailbox is full, so I'll post in on RCM.


<rdsandman[remove]comcast.net> wrote:

[snip]

assumption is that they would have firearms in the home. If the folks had not had CCW records on file, I wonder what the participation would have been.

specifically selected *because* they are CCW permit holders.

And from whom were their names obtained?
The government. They knew lying would not keep them off any list of gunowners they were already on.

Some of you seem to have mistaken how this study was done. First, it wasn't a unique study. It was a seeded test *within* the 2001 National Gun Policy Survey (NGPS).
No one except the test designers knew who the CCW holders were. No one except them even knew there *were* CCW holders seeded among the random sample. The interviewers didn't know and the interviewees didn't know.
This was a completely double-blind test. The people who held CCW permits had no reason to believe they had been chosen because they had carry permits -- and no reason to know that anyone involved with the research even knew they had them.
So the results are pretty persuasive.
I have the complete study in PDF format, if you have any questions about it.
I'd be interested. Thanks,
Joe
Email address is good.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

Sorry, Joe. The mailbox is full because it doesn't exist. <g> My apology. I have to start again to indicate my real address. Delete the "3" from the one you have. That's my real address.

Ok. Off it goes...
Ed Huntress
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

OK

Got it. Thanks. Will read.
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

I was beginning to wonder.

I've now read it, and have a few observations:
The stated rationale for making this study is to show that the results of such surveys are correct enough to be used in policymaking.
The claim is that by seeding the sample with CCW holders, they are able to tell if people on average tell the truth about their gun use and possession to telephone interviewers who cold call.
Now surveys that poll only a subset of the population are required to demonstrate that the sample is representative of the larger population not polled before one may draw conclusions about the population from such a sample. This is why everything possible is randomized - true random sampling increases the error band (margin of error) but does not bias the answer one way or the other.
In other words, the sample must be truly representative of the often far larger population it claims to represent.
So right away, we have an element of circularity. We are using the CCW holders (a subsample of the sample) to estimate if the sample at large tells the truth to interviewers. But CCW holders are already known to tell the truth to a government (never mind an interviewer), and so cannot be used to deduce the behavior of non-CCW holders. They are not a random subset. In other words, this methodology begs the question.
This is a methodological flaw that invalidates that part of the study.
Also, on page 443 it was noted that of 778 CCW holders in the original contact list, there were 300 interviews (39%), 230 refusals (30%), and 194 people not reached (25%). In round numbers, almost as many refused to be interviewed as agreed. I don't recall seeing a similar analysis for the larger full sample (it should be in the referenced studies), but one would hazard that the refusal rate in the larger sample must be similar. (This assumes that the CCW holders are a random subset of the entire sample, which they aren't, but never mind.)
The problem here is that if more gun owners refuse than non gun owners, the effective sample (those answering) will be enriched with non gun owners, thus biasing the study in favor of non gun owners. It is impossible from the published data to estimate how large this bias is or might be.
Recalling the discussion about the apparent decline in gun ownership and the change in texture of the curve in 1988, another issue comes to mind - the politics changed.
If in the 1960s and 1970s you had asked my parents if they possessed and used firearms, they would have thought nothing of answering truthfully.
Now days, I'm not so sure. My father later obtained a CCW permit, but did not carry. Nor was he worried about self defense. His reason to get a CCW was that the gun laws in Mass had gotten so strange that he was afraid of making a mistake and getting nailed. A CCW eliminated that risk.
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:

Since you're being precise here, the CCW permit holders are not a subsample. They're a new sample, chosen from the whole body of CCW holders, and they were added to the original random sample of people. In the lingo of polling, they are "seeds."

Sort of. The researchers were qualifying their results a bit more restrictively, but that's the ideal version of what they would like to have measured.

My opinion is that you're asking too much of it, and you're trying to draw too much from it. The researchers knew the limitations of what they were trying to do. Basically, the idea was to get a measure of how truthful people who were *known* to have guns (a CCW permit holder without a gun is a joke made up by someone in this thread), about the guns they are known to have. And they got that result.
Anytime you take some modest result like that and try to project it, one has to ask what it is you're trying to do with this limited information, and what you would do about it if you had *perfect* information.
I think that you have some very useful information from that study. But it doesn't matter how perfect it is, if what you want to know is how people's truthfulness changes over time. This is a snapshot. If you think there is some significant change in truthfulness over time, you'd better design and conduct another study. This one tells you about CCW holders in 2001.
As for public policy, now they don't have to contend with some naysayer who would claim "but half of those people wouldn't be truthful." Now they have a better answer. Trying to narrow it down much more than that is foolish, but knowing that the large majority of people in this study told the truth provides enough confidence to do other studies with a reasonable degree you're not wasting your time.
A lot of policy-making is done that way. A lot of *business policy* decision making is done that way. You go with the probabilities, and you hope that you win more than you lose. If you have a real, practical, and mature sense of statistical studies, you probably will.

Again, without this study to give us an idea of whether the "non-gun owners" are massive liars, you could have a problem. But now you have somthing that indicates that gun owners are not disproportionate liars. That's what the seeded sample does for you.

I don't see how the "texture" of the curve could possibly influence the result. It was an up-and-down bobbling, within the MOE.

All in all, combined with the fact that we know we have over 14 million fewer hunters than we would have if the percentages hadn't changed since 1960, it all points to the idea that we have fewer gun owners with more guns. You can pick at the statistics but at some point you have to come down somewhere, and the evidence comes down on the side of the idea there are fewer gun owners.
From my own experience, it agrees with what I've seen over my lifetime. But I'm not counting on anecdotal experience to draw a conclusion.
--
Ed Huntress


>
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Yes, I know. That actually makes it worse - the subsample is even more different than the sample it is to represent.

Yes, that is what they were trying to measure. The question is if they found a tool capable of measuring any such thing.
Begging the question is a pretty fundamental problem.
.<https://en.wikipedia.org/wiki/Begging_the_question > Anytime you take some modest result like that and try to project it,

Well, the authors' stated purpose was to inform policymaking, so it is worthwhile to figure out if their study has any ability to answer the question they posed.
Otherwise, we run the risk of misinforming those policymakers.

Yes. It tells one about CCW holders in 2001, and about a different population of largely non CCW holders, but cannot tell you how to compare data from one group with the other.

I'm not convinced that this can silence the naysayers, for the reasons we are discussing.

In other words, we go with gut instinct, not data? The problem with a "mature sense of statistical studies" is that it is subjective (we all believe that we are the mature one), which puts us right back where we were before the studies were done.

Hmm. Massive mendacity is not required.
It wouldn't take extraordinary mendacity to tilt the results enough to explain the results. Nor would it take much refusal to be interviewed to cause sufficient enrichment of the sample, and we know we have both effects, but don't have any way to tell how widespread this is. And this study cannot answer the question.
People do tend to be economical with the truth when polled. I recall an amused article from maybe 30 years ago on estimating magazine circulation by telephone polling people: The polls showed that magazines like Time and Newsweek and The Atlantic were the top sellers, but audited circulation figures told a very different story: the tabloid scandal magazines outsold these upstanding publications by a factor of ten. (You probably have more recent examples.)
The extreme example would be a survey trying to estimate how many people use illegal drugs by calling them up on the phone and asking. It's only a felony.

The change in plot texture is significant and quite apparent to the eye. The question is what caused the change.
My prior speculation was that Gallop changed its polling methodology.
To that I now add the observation that the politics had also changed, arguably in such a way as to reduce the effectiveness of polling by biasing it.

I have no opinion on the truth of this claim.
But even if it is true, is it important? We still have no shortage of guns. And tens of millions of new guns are sold every year.
But as I said before, it is unclear what difference it makes if there are 400 million guns, or merely 200 million, rising or falling. It's still of order one gun per person (every man, woman, and child), and will remain so for the foreseeable future.
Hmm. Assume that we have only 200 million guns. The FBI reports that 20 million guns were sold in 2014, or 10%. It seems quite unlikely that 10% of the gun inventory was melted down or otherwise lost in that year, so the growth rate is about 10% per year, and it won't be long before 400 million is surpassed.
.<https://www.fbi.gov/about-us/cjis/nics/reports > From my own experience, it agrees with what I've seen over my

Well, yes. But we do need better-designed studies. But there are questions that cannot be answered by polling, and we will just have to live with that.
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
wrote:
Mr. Ed said:

Fewer gun owners? This I doubt. I've talked to several of my old friends recently and found that they either got CCWs or new handguns recently. The reasons given: they saw data on the sparse ratio of cops to people, media fear-mongering (despite the dropping crime rate), fear of gun sale bans, etc.

And, just as they undercount illegal aliens, I'm sure they miss lots of guns and gun owners in their hunts.

Ayup.

Agreed. Polls can be counted on to give some semblance of the truth, not the whole truth, EVER.
If Ed ever gives up his hard-on against the NRA, he may see that.
--
My desire to be well-informed is currently
at odds with my desire to remain sane. --Sipkess
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload


The people involved in polling have a strong financial interest in convincing customers of their value.
http://www.telegraph.co.uk/news/general-election-2015/11457970/If-you-want-to-know-who-will-win-the-election-ask-a-gambler.html
-jsw
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sat, 25 Jul 2015 05:51:09 -0700, Larry Jaques

Aha, Dr. Demento sticks his oar in...
Larry, you've now contradicted yourself within the span of three sentences -- a new record!
(If you doubt that there are fewer gun owners, and you base it on the fact that *existing* gun owners got CCWs or new guns, you've just jumped beyond a nonsensical conclusion, and landed flatly on top of your head.)
--
Ed Huntress

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.