|
Cellular forums Home > Archive > AT-T wireless service > March 2007 > Re: Steven's Myth of Verizon AMPS coverage in the San Francisco Bay Area
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Re: Steven's Myth of Verizon AMPS coverage in the San Francisco Bay Area
|
|
| Todd Allcock 2007-02-15, 10:33 pm |
| At 15 Feb 2007 17:23:52 +0000 John Navas wrote:
> The good reason is that CR subscribers have considerably different
> demographics than the universe of cellular, making any CR survey
> unrepresentative.
So that begs the question is the CR subscribers' demographic more likely
or less likely to have good service or bad service? I can't personally
envision a scenario that would cause CR subscribers to have vastly
different cell service than non-subscribers.
> Then there's the problem of self-selction.
I'll agree that's a bigger problem, assuming that unhappy customers might
be more likely to "grumble" about service than happy ones are to gush.
But again, that would drag everyone's scores down- not one carrier's, so
the relative results wouldn't be much different.
I suppose the easiest way to make the study accurate for me, would be
simply to subscribe to CR, then the results would apply to me, since I
would be part of the represented universe.
I'll let you all know how my subscription to CR affects my cellphone
reception... ;-)
| |
| John Navas 2007-02-15, 10:33 pm |
| On Thu, 15 Feb 2007 14:15:19 -0700, Todd Allcock
< elecconnec@AmericaOn
Line.com> wrote in <er2lbh$drb$3@aioe.org>:
>At 15 Feb 2007 17:23:52 +0000 John Navas wrote:
>
>
>So that begs the question is the CR subscribers' demographic more likely
>or less likely to have good service or bad service?
With all due respect, there is no such question -- you can only make
valid generalizations from proper sampling, and the CR survey isn't even
a proper sampling of CR subscribers (because of self-selection).
There's simply no way to know the sample bias. It's thus just pure
speculation.
>I can't personally
>envision a scenario that would cause CR subscribers to have vastly
>different cell service than non-subscribers.
I can easily think of lots of things that would bias the result in
unpredictable ways, but without real data, there's simply no way of
knowing which of them might be at work.
>
>I'll agree that's a bigger problem, assuming that unhappy customers might
>be more likely to "grumble" about service than happy ones are to gush.
>
>But again, that would drag everyone's scores down- not one carrier's, so
>the relative results wouldn't be much different.
There are lots of other possible factors that might well differ from
carrier to carrier, but again, without real data it's all just
speculation.
>I suppose the easiest way to make the study accurate for me, would be
>simply to subscribe to CR, then the results would apply to me, since I
>would be part of the represented universe.
1. That's not how statistics works. The issue is learning something
about the entire target universe, not any one individual.
2. That wouldn't solve the problem of self-selection.
>I'll let you all know how my subscription to CR affects my cellphone
>reception... ;-)
Whatever.
--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/ Cingu...less_FA
Q>
| |
| John Navas 2007-02-15, 10:33 pm |
| On Thu, 15 Feb 2007 16:09:59 -0800, SMS <scharf.steven@geemail.com>
wrote in < 45d4f65d$0$27188$742
ec2ed@news.sonic.net>:
>Todd Allcock wrote:
>
>
>Even though there are certainly some demographic differences between CR
>subscribers and the general public, none of these would materially
>affect the results of one carrier more than another carrier.
There's simply no way of knowing that without real data, which doesn't
exist, so that's just your own unsupported speculation.
>The
>demographics of Consumer Reports readers versus the general population
>include higher education level, higher income, more moderate
>politically, and middle-age to older age.
Citation? Apparently not.
>On the one hand, these
>demographics would tend to make people more critical of products and
>services, but on the other hand, these demographics may understand the
>limitations of cellular communications better than uneducated people, so
>they may cut the carriers more slack.
Again, just your speculation.
>What makes the CR survey so valuable is that they don't just ask "how's
>your service?" they go into great detail with questions on specific
>aspects of the service.
Actually, it doesn't -- the survey is based on simple generic multiple
choice questions (as you should know if you had any actual knowledge of
the methodology).
>But as you stated, even if the people that
>choose to respond are more or less critical of a service, this will
>extend across all the carriers, and will cancel out.
Again, just your speculation.
>In earlier years, Navas complained that the reason Cingular was rated so
>poorly was that their TDMA subscribers were dragging down their scores
>(though in reality the TDMA/AMPS subscribers were probably dragging the
>scores up).
What I actually did was point out the invalidity of lumping multiple
types of GSM and TDMA together, just like the invalidity of lumping
together Sprint CDMA2000 and Nextel iDEN.
>Now that the TDMA subscribers are an insignificant portion
>of the total, he's come up with a new excuse, apparently he believes
>that the Cingular subscribers that choose to respond to the survey are
>somehow more critical of their service than the Verizon subscribers that
>choose to respond.
I've said nothing of the kind. What I have done is point out the
invalidity of the CR survey. Regardless, as I expected, Cingular is
doing much better indeed in the market now that most of the migrations
are behind it.
>Certainly if there were any evidence that CR's
>subscribers demographics somehow benefit one carrier over another, he'd
>have presented that evidence, or at least a theory of that evidence.
On the contrary -- there's simply no way to know the sampling bias
without real data, which simply doesn't exist.
>As
>you pointed out "non-random and non-representative are not the same thing."
I guess you have no idea how meaningless that is.
--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/ Cingu...less_FA
Q>
| |
| Scott 2007-02-15, 10:33 pm |
| John Navas < spamfilter1@navasgro
up.com> wrote in
news:aqn9t294p65unfp
3hpt9maatd4oi5p58eo@
4ax.com:
> On Thu, 15 Feb 2007 14:15:19 -0700, Todd Allcock
> < elecconnec@AmericaOn
Line.com> wrote in <er2lbh$drb$3@aioe.org>:
>
likely[color=darkred
]
>
> With all due respect, there is no such question -- you can only make
> valid generalizations from proper sampling, and the CR survey isn't
even
> a proper sampling of CR subscribers (because of self-selection).
> There's simply no way to know the sample bias. It's thus just pure
> speculation.
So you are saying that you can't prove any sample bias. Thanks for
playing.
>
>
> I can easily think of lots of things that would bias the result in
> unpredictable ways, but without real data, there's simply no way of
> knowing which of them might be at work.
Bullshit. Unless you can prove that CR subscribers have a skewed view
of service and coverage, it is a very real-world sample.
>
might[color=darkred]
so[color=darkred]
>
> There are lots of other possible factors that might well differ from
> carrier to carrier, but again, without real data it's all just
> speculation.
>
You keep talking about "other factors"- there are none. Either they
have the same view of service or they don't. Period.
>
> 1. That's not how statistics works. The issue is learning something
> about the entire target universe, not any one individual.
>
> 2. That wouldn't solve the problem of self-selection.
All surveys are self-selected. No survey is done where the respondent
is required to answer questions. All those surveyed make the conscious
decision to participate in the survey, and therefore self-select.
>
>
> Whatever.
>
What? He can't use your argument against you?
| |
| Scott 2007-02-15, 10:33 pm |
| SMS <scharf.steven@geemail.com> wrote in
news:45d4f65d$0$2718
8$742ec2ed@news.sonic.net:
>
> I read a funny article about that in SmartMoney magazine yesterday. It
> seems that there is very little middle ground in most reviews by
> individuals of products and services. They either hate something or
> love it. I.e. a five star review for a stapler at Amazon, "This
> stapler is great. It works very well and staples many papers
> together."
>
As someone responbile for overseeing the Customer Satisfaction program
(surveys included) for a division a Fortune 100 company, I can see many valid
points in this. Very few surveys are done where a middle-of-the-road rating
is given. The majority of responses are typically the highest and lowest
rating available on the survey.
| |
| John Navas 2007-02-15, 10:33 pm |
| On Thu, 15 Feb 2007 17:25:40 -0800, SMS <scharf.steven@geemail.com>
wrote in < 45d5081a$0$27225$742
ec2ed@news.sonic.net>:
>Scott wrote:
>
>Bingo!
That's not what self-selected means. Learn something about statistics.
--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/ Cingu...less_FA
Q>
| |
| Scott 2007-02-15, 10:33 pm |
| John Navas < spamfilter1@navasgro
up.com> wrote in
news:1h2at2pvrdkngqu
liu18miur324raej3td@
4ax.com:
> On Thu, 15 Feb 2007 17:25:40 -0800, SMS <scharf.steven@geemail.com>
> wrote in < 45d5081a$0$27225$742
ec2ed@news.sonic.net>:
>
>
>
> That's not what self-selected means. Learn something about statistics.
>
Sorry, John- I use more statistics in the first hour of my day than you
have ever been able to glean from your Google searches.
| |
| John Navas 2007-02-16, 12:33 pm |
| On Fri, 16 Feb 2007 08:50:27 -0800, SMS <scharf.steven@geemail.com>
wrote in < 45d5e0d9$0$27255$742
ec2ed@news.sonic.net>:
>Scott wrote:
>
>
>According to The Journal of Consumer Research, Vol. 4, No. 4 (Mar.,
>1978), pp. 247-251, "The Consumer Reports subscriber is found to be
>richer, better educated, and more likely to own a fairly wide range of
>durable goods."
>
>Given the demographics, it'd be interesting to know whether the CR
>subscriber base, which is better educated and more affluent than the
>general public, is more or less critical of goods and services.
>
>In any case, any difference cancels out because it would be present no
>matter which carrier the subscriber was reviewing.
Statistics simply doesn't work that way:
* There's no way to know the bias without data that doesn't exist.
* Bias can only be eliminated by proper sampling techniques.
>CR surveys are
>designed to eliminate this sort of bias, so any differences between the
>CR subscriber base and the general population cancel out.
That's simply untrue.
--
Best regards, FAQ FOR CINGULAR WIRELESS:
John Navas <http://en.wikibooks.org/wiki/ Cingu...less_FA
Q>
| |
| clifto 2007-02-16, 10:33 pm |
| Scott wrote:
> Sorry, John- I use more statistics in the first hour of my day than you
> have ever been able to glean from your Google searches.
A-ha! An actuary! http://users.aol.com/fcas/jokes.html
--
"Nowadays, security guys break the Mac every single day. Every single day,
they come out with a total exploit, your machine can be taken over totally.
I dare anybody to do that once a month on the Windows machine."
-- Bill Gates, in an interview with Newsweek's Steven Levy
| |
| Todd Allcock 2007-02-16, 10:33 pm |
| At 16 Feb 2007 17:28:11 +0000 John Navas wrote:
the[color=darkred]
>
> That's simply untrue.
Yet also irrelevant. The survey can stand alone as a specific population-
i.e. "4 out of 5 dentists."
Like it or not, "a survey of CR subscribers found brand X better than Y"
carries weight as a data point. And your statistics 101 excuses for it's
lack of validity might be a reason to keep it out of your doctoral
thesis, but not a reason to invalidate it's value as a reference for
people researching cell service. In many ways, a limited population
survey is MORE useful. For example, I would prefer a survey of "best
luggage" to be comprised only of, say, frequent business travelers, or
airline pilots, over one culled from the "general population," many of
whom don't travel or travel infrequently.
| |
|
|
|
|
|
|
|
|
|
|
|
|
|