Extended Warrenty is it worth it , have you gotten your moneys worth out of it ?

Take the money and put it into a money market account or short term
cd. Only take money out to pay covered items less deductible. when
you hit the time or mileage that cancels the policy take the remaining
money and have a party. I bet you will have a good one.

Had a friend once who was the top appliance salesman at a major
retailer. He got fired. Know why. He didn't push extended
warranties hard enough and wasn't getting his quota. You can see
where the money is being made.

On Mon, 18 Jul 2005 11:08:12 GMT, Mr X > wrote:

>
>I declined the dealer offer of a 5yr 100k mile extended warrenty,
>first the $ was 2900 then at the end it was 1600 with a $100
>deductable and I still declined . I may have made a mistake ??? time
>will tell
>
>
>Does chrysler give it another shot at selling me a warrenty in a month
>or so thats what some have told me? I got a 2005 sebring with ALL the
>discounts,
>
> note my other car is a 89 ciera with 197,000 miles on it that I
>bought new. I do most of the work on it myself with the exception of
>mufflers and front suspension parts, I never had to do transmission
>or internal engine repair on it and would pass on that. I'm a
>electronic tech and plan to get a computer interface scan tool for the
>Sebring.
>
>
>thanks for any reply

Frank wrote:
> >So, I guess you are right, On average Chrysler reliability is pretty much
> >mediocre. As I was looking through the CR information I was surprised to see
> >that in almost ever category, the Chrysler products were from the middle to
> >near the bottom of the vehicle ratings. Apparently CR is not a big Chrysler
> >fan.
> >
> >Ed

Consumer Reports is just the compilation agency here. Those reports are
what they compile from the rather large number of people who fill out
the survey questionnaires. It has nothing to do with the editors or hte
mag per se unless they lie and fudge the numbers. Since several hundred
thousand questionnaires are completed, a huge number, the results are
somewhat accurate even if random sampling was not used - most likely
self-selected sampling. It's also possible that the several hundred
thousand people dislike Chrysler, but unlikely.

I've owned mostly Chrysler products and I would often compare my
experiences with those CR reports. They were dead on. But at least I
knew what was going to break, in the general case

The only time I noticed bias with CR is when they did a real
double-talk number about large cars versus small cars and safety. What
utter nonsense! I was embarrassed by their twisting physics and
statistics. For the record, when a large vehicle crashes into a small
vehicle, it's not good for the small vehicle - as a general rule of
thumb. Even CR cannot twist the laws of nature.

wrote:
> Frank wrote:
>
>>>So, I guess you are right, On average Chrysler reliability is pretty much
>>>mediocre. As I was looking through the CR information I was surprised to see
>>>that in almost ever category, the Chrysler products were from the middle to
>>>near the bottom of the vehicle ratings. Apparently CR is not a big Chrysler
>>>fan.
>>>
>>>Ed
>
>
> Consumer Reports is just the compilation agency here. Those reports are
> what they compile from the rather large number of people who fill out
> the survey questionnaires. It has nothing to do with the editors or hte
> mag per se unless they lie and fudge the numbers. Since several hundred
> thousand questionnaires are completed, a huge number, the results are
> somewhat accurate even if random sampling was not used - most likely
> self-selected sampling. It's also possible that the several hundred
> thousand people dislike Chrysler, but unlikely.

Yes, but they don't have a random sample of the general population. I
once saw the demographics of they subscriber base and it most decidedly
did not reflect the general American public.


Matt

Matt Whiting wrote:
> > Consumer Reports is just the compilation agency here. Those reports are
> > what they compile from the rather large number of people who fill out
> > the survey questionnaires. It has nothing to do with the editors or hte
> > mag per se unless they lie and fudge the numbers. Since several hundred
> > thousand questionnaires are completed, a huge number, the results are
> > somewhat accurate even if random sampling was not used - most likely
> > self-selected sampling. It's also possible that the several hundred
> > thousand people dislike Chrysler, but unlikely.
>
> Yes, but they don't have a random sample of the general population. I
> once saw the demographics of they subscriber base and it most decidedly
> did not reflect the general American public.

Is that not what I was saying, apparently not too well.

I said, "random sampling was not used" - which means just that, not
used.

But if you get a huge non-random sample, then it is possible to
approach the benefits of random sampling. In other words, the best
sample is not random, but all, that is, the entire universe, of all
people who bought Chrysler products!

When you have everybody, then there is not any bias possible.

Let's see, if I remember the Central Limit Theorem, if the numbers are
big enough, the bias falls aways and it's not required to be random.
Random sampling is used mostly when the sampling is small because of
money reasons.

You say the demographics of the subscriber base did not reflect the
general American public. That's for sure. CR's base is most likely much
higher in education and earned income, that is, socioeconomic status,
compared to the average American, and most likely far more liberal
than, oh, the Bible Belt.

But how does that affect whether a muffler fails or an engine needs
repairing. Do you think they lie when they fill out the questionnaires?

Do you think Chrysler products behave better when a Bible Belt person
with less education is driving the product as opposed to a liberal
northerner?

Just asking

wrote:
> Matt Whiting wrote:
>
>>>Consumer Reports is just the compilation agency here. Those reports are
>>>what they compile from the rather large number of people who fill out
>>>the survey questionnaires. It has nothing to do with the editors or hte
>>>mag per se unless they lie and fudge the numbers. Since several hundred
>>>thousand questionnaires are completed, a huge number, the results are
>>>somewhat accurate even if random sampling was not used - most likely
>>>self-selected sampling. It's also possible that the several hundred
>>>thousand people dislike Chrysler, but unlikely.
>>
>>Yes, but they don't have a random sample of the general population. I
>>once saw the demographics of they subscriber base and it most decidedly
>>did not reflect the general American public.
>
>
> Is that not what I was saying, apparently not too well.
>
> I said, "random sampling was not used" - which means just that, not
> used.

Yes, but you also said "somewhat accurate" and that isn't the case at
all with their technique. There results may be accurate randomly, but
certainly not consistently.


> But if you get a huge non-random sample, then it is possible to
> approach the benefits of random sampling. In other words, the best
> sample is not random, but all, that is, the entire universe, of all
> people who bought Chrysler products!

Actually, a large non-random sample doesn't at all approach the benefits
of a random sample. It just gives you a higher confidence level in your
inaccurate result!


> When you have everybody, then there is not any bias possible.

When you have everybody you don't have a sample, you have the population.


> Let's see, if I remember the Central Limit Theorem, if the numbers are
> big enough, the bias falls aways and it's not required to be random.
> Random sampling is used mostly when the sampling is small because of
> money reasons.
>
> You say the demographics of the subscriber base did not reflect the
> general American public. That's for sure. CR's base is most likely much
> higher in education and earned income, that is, socioeconomic status,
> compared to the average American, and most likely far more liberal
> than, oh, the Bible Belt.
>
> But how does that affect whether a muffler fails or an engine needs
> repairing. Do you think they lie when they fill out the questionnaires?

Lie is a strong word, but I've seen many that perceive things
differently. I've known people who own both American brands and Foreign
brands. They complain about every little thing on the American car and
seem to ignore similar problems on their imports as though it was just a
fluke, as opposed to being a "design flaw" on the American car. I don't
claim to understand the psychology of it, but I've witnessed it many times.


> Do you think Chrysler products behave better when a Bible Belt person
> with less education is driving the product as opposed to a liberal
> northerner?

Probably, because people with less education seem to take better care of
their cars! I have a couple of college degrees and I can tell you that
few of my colleagues maintain their own cars or even pay much attention
to when the maintenance gets done. I grew up in a poor part of
northcentral PA and everyone I knew did their own maintenance and knew
how to do most of their own repairs. They had too as they simply
couldn't afford to run to the garage every time something broke. So,
yes, I DO think cars are treated differently by different groups of owners.


> Just asking

Just answering. :-)

Matt

Matt Whiting wrote:
> wrote:
> > Matt Whiting wrote:

> >>>Consumer Reports is just the compilation agency here. Those reports are
> >>>what they compile from the rather large number of people who fill out
> >>>the survey questionnaires. It has nothing to do with the editors or hte
> >>>mag per se unless they lie and fudge the numbers. Since several hundred
> >>>thousand questionnaires are completed, a huge number, the results are
> >>>somewhat accurate even if random sampling was not used - most likely
> >>>self-selected sampling. It's also possible that the several hundred
> >>>thousand people dislike Chrysler, but unlikely.

> >>Yes, but they don't have a random sample of the general population. I
> >>once saw the demographics of they subscriber base and it most decidedly
> >>did not reflect the general American public.

That's true but irrelevant, if I may use so harsh a word , because we
are describing cars, not the general American public! This ain't no
voter survey, it's a car survey.

And the numbers for any one specific vehicle are way less than the
number of Americans in the USA.

I know, you'll say the bias of the subject pool influences their
reponses. And I'll say, you're right, but the bias is not enough to
offset the fact that Japanese cars are better made than American cars -
any place for me to duck now? And they were using American methods
[forgot that quality control guy, an American, who went over there when
Japan could not make anything right] originally although now we are
using Japanese methods. Go figure!

> > Is that not what I was saying, apparently not too well.

> > I said, "random sampling was not used" - which means just that, not
> > used.

> Yes, but you also said "somewhat accurate" and that isn't the case at
> all with their technique. There results may be accurate randomly, but
> certainly not consistently.

To be accurate randomly, by definition, is to be inconsistent. Mais
oui?

Are you suggesting a bias? If so, that would make the results quite
consistent, but consistently biased or wrong. Bias is not a good thing
here.

I think this touches on the potential problem of
/heteroscedasticity/homeoscedasticity which means the standard errors
are or are not randomly distributed? If the errors were random or
homeoscedastistic then the sampling was most likely good, even if not
random. Is that so? Do you remember statistical inference? I'm trying
to.


> > But if you get a huge non-random sample, then it is possible to
> > approach the benefits of random sampling. In other words, the best
> > sample is not random, but all, that is, the entire universe, of all
> > people who bought Chrysler products!
>
> Actually, a large non-random sample doesn't at all approach the benefits
> of a random sample. It just gives you a higher confidence level in your
> inaccurate result!

Any large sample increases the possibilities of statistical
significance. I guess that's what you meant by a higher confidence
level? Sure it is. I see your point but I am not sure about it. If I
work out the numbers, then I was wrong to suggest that the large
non-random sample might overcome even the random sample for accuracy. I
doubt if Consumer Reports samples that high.

This is a problem with statistical inference in general and has nothing
directly to do with random or non-random sampling. As n or the number
of subjects goes sky high, so does the probability of statistical
significance.

Let me go back to the problem of n or number of subjects and random or
non-random.

Generally most polls, like Harris or Gallup, use about 3,000 subjects
to handle the USA. According to stat theory, 1,200 with the normal
distribution curve, should be enough. So why the over-sampling? It's to
make sure the random gets enough since it's quite possible if you play
it close to the bone, you miss something. Now I see stratified,
representative random sampling. They make sure to random sample those
who might get missed. This is a huge problem with the Census Bureau. In
any way, 3,000 people done by random sampling, with stratification and
representation worked in, do get the pulse of the huge quarter of a
billion people in the USA. Yes?

CR uses a quarter of a million questionnaires. That would be great if
it were only for one specific Chrysler car. But they do what, 100 cars,
including trucks and foreign cars? So how is that broken down by
specific vehicles. If it's about 3,000 questionnaires for a specific
vehicle, I would say they are in the ball park, but that's way too
close for comfort for what is needed for just even a random sample. But
it may be adequate for a particular car where the universe for that car
is what, at most, 300,000 vehicles? A million seems too high for a
specific model but even for a million vehicle, 3,000 non-random
questionnaire might be doable. We are not trying to approximate the
entire USA population. We are trying to approximate the entire
population of one particular vehicle. I think here is a possible
mistake that you were making in your deconstruction of the random
sampling versus CR.

So I would agree with you here about your comments on random and
non-random sampling from a theoretical point of view, except for the
previous caveats. It's good you point that out. But I don't think the
lack of random sampling biases the results beyond use. And I do think
the reason is the large quantity of asking. When you view their CR
results with that of auto experts and magazines, there's not much
difference, so that's another big plus for CR.

However from a practical point of view, CR has always been on the money
for me. My bibles when going to buy a used car were old CR annual books
which listed all the cars for the previous 7 years or so. I still
bought Chrysler cars but now I knew what was going to break!

I've owned Chrysler cars except for 1 Oldsmobile and 1 Toyota Tercel.
And CR was dead on for those cars. Nothing random or inaccurate about
their trouble spots for those cars. In fact, I used to say I was
embarrassed as an American by how well the Toyota TErcel was
engineered, given that I had the cheapest, barest, ugliest Toyota made!
But it was damn well made. "Intelligent design" is how I would put it.

I have noticed the trend for people to ignore the problems of their
Japanese cars but that's because they do break less than all others,
even the Europeans. Women know this and that's why they buy Japanese
cars. They don't like the hassles of dealing with male mechanics. I
don't like the hassles of dealing with many male mechanics who are not
all that truthful and polite. They are as bad as doctors or quacks.

Given all that, my current car is within a thousand miles of 200,000
miles on the odometer. It does not leak or use oil. It needs an oxygen
sensor and maybe a new tire or axle and it's a Chrysler Plymouth
Voyager 1994 with the infamous 3.0 liter Mitsubihi engine.


> > Just asking
>
> Just answering. :-)
>
> Matt

Thanks for your replies.

> wrote in message
oups.com...

> That's true but irrelevant, if I may use so harsh a word , because we
> are describing cars, not the general American public! This ain't no
> voter survey, it's a car survey.
>

And your ignoring the entire principle of why a self-selected survey is
bogus.

The people who are driving vehicles with no problems have little incentive
to return a survey and so toss it. The people with vehicles that have lots
of
problems - either due to the vehicle flaw or their own stupidity in
maintaining
the vehicle - are mad about it and have an axe to grind and are very happy
to fill out a survey to "get even with the ****ter who made my POS car"

There's more domestics than imports so if the same pecentages of vehicles
in each group have problems, then there's more quantity of problem
domestics than problem imports, so you get more surveys bitching about
problem domestics.

> And the numbers for any one specific vehicle are way less than the
> number of Americans in the USA.
>
> I know, you'll say the bias of the subject pool influences their
> reponses. And I'll say, you're right, but the bias is not enough to
> offset the fact that Japanese cars are better made than American cars -
> any place for me to duck now? And they were using American methods
> [forgot that quality control guy, an American, who went over there when
> Japan could not make anything right] originally although now we are
> using Japanese methods. Go figure!
>

The only survey that would have meaning is a satisfaction survey that
polled the numbers of people happy with their cars in each group, then
normalized for market share/number of cars sold/geographic distribution
of vehicles sold/etc/etc/etc. But then, you would have a scientific
survey, which would have real meaning, but would not allow you to
make up flashy headlines that sell magazines.

If the majority of domestic car owners that are happy with their cars
see a magazine title in the rack that says "owner surveys say imports
are pieces of ****" why are they going to buy it, they already know that.
However, if the title says "owner survey says domestics are POS"
then the domestic car owner thinks "Damn, that's bull****, what are
those peckers talking about" and picks up the ragazine to look at
it. Since theres more domestic car owners than import car owners,
you have more magazine sales if you wank off the majority, rather than
the minority.

> > > Is that not what I was saying, apparently not too well.
>
> > > I said, "random sampling was not used" - which means just that, not
> > > used.
>
> > Yes, but you also said "somewhat accurate" and that isn't the case at
> > all with their technique. There results may be accurate randomly, but
> > certainly not consistently.
>


> To be accurate randomly, by definition, is to be inconsistent. Mais
> oui?
>
> Are you suggesting a bias? If so, that would make the results quite
> consistent, but consistently biased or wrong. Bias is not a good thing
> here.
>
> I think this touches on the potential problem of
> /heteroscedasticity/homeoscedasticity which means the standard errors
> are or are not randomly distributed? If the errors were random or
> homeoscedastistic then the sampling was most likely good, even if not
> random. Is that so? Do you remember statistical inference? I'm trying
> to.
>

All of this is bull**** because it is very easy to get a large group of
humans to believe the WRONG thing. For example, most Americans
believed that we were going into Iraq because of WMDs, not because
of oil. That didn't make it any truer.

If you are talking about sampling apples or something, then your theorm
applies, but it goes out the window on a survey of people.

Most people believed the world was flat one time and a giant study
based on that would have proven it was flat. Fortunately, even back
then, enough intelligent people knew this argument was bull**** and
so discovered the New World.

>
> Generally most polls, like Harris or Gallup, use about 3,000 subjects
> to handle the USA. According to stat theory, 1,200 with the normal
> distribution curve, should be enough. So why the over-sampling? It's to
> make sure the random gets enough since it's quite possible if you play
> it close to the bone, you miss something.

That isn't why they oversample. They oversample because they are
telephoning each subject, and since some people don't answer the phone,
you aren't guarenteed that the number of people who are NOT responding
is randomly distributed. Most of the time it isn't, so you have to
oversample
so you can toss out some respondents in order to compensate for the
nonrandom distribution of the nonrespondents.

> entire USA population. We are trying to approximate the entire
> population of one particular vehicle. I think here is a possible
> mistake that you were making in your deconstruction of the random
> sampling versus CR.
>

And you are just wrong, you cannot use the kinds of statistical numbers
arguments that apply to sampling manufactured goods on an assembly
line to apply to sampling people.

> So I would agree with you here about your comments on random and
> non-random sampling from a theoretical point of view, except for the
> previous caveats. It's good you point that out. But I don't think the
> lack of random sampling biases the results beyond use. And I do think
> the reason is the large quantity of asking. When you view their CR
> results with that of auto experts and magazines, there's not much
> difference, so that's another big plus for CR.
>

All of those auto experts and magazines make their money by selling
their results, and nobody buys results that tell them what they already
know. They buy results that seem different than what Real Life tells them
because they are curious to understand why it is that a survey tells
them that what they know from direct exposure is wrong. Thus, there
is ALWAYS a strong financial motivation among the peddlers of
these surveys to skew the results to be controversial, so that they
make more money selling surveys.

I see this in business all of the time. About once a quarter in some
meeting a marketing guy will drag out some expensive survey that
he wasted the companies money on, claiming that some new, unknown
product that nobody ever thought that they needed or wanted before,
is now poised to become the next big seller and we gotta catch the
wave now. Of course, the survey peddler gussies up the survey with
colorful graphs and expensive bindings, but it still is nothing more
than a product designed to sell copies of itself.

The ONLY surveys that have any validity whatsoever are the ones
that get published in a scientific journal for someone's doctorate
thesis, or some such, because those people do not make money by
selling more and more thesis copies, and those surveys are subject
to exhaustive scientific review and must withstand many challenges.
They also must make the sources the surveys were compiled from,
available.

The CR car surveys, like most industry surverys, are subject to NO
outside peer review. CR doesen't even bother to print survey
rebuttals, to give the users opposing viewpoints. They are
entertainment value only.

> However from a practical point of view, CR has always been on the money
> for me. My bibles when going to buy a used car were old CR annual books
> which listed all the cars for the previous 7 years or so. I still
> bought Chrysler cars but now I knew what was going to break!
>
> I've owned Chrysler cars except for 1 Oldsmobile and 1 Toyota Tercel.
> And CR was dead on for those cars. Nothing random or inaccurate about
> their trouble spots for those cars.

It's just selective memory.

CR says 4 items on a GV are trouble prone, the trans, the wipers, the
ac, and the floormats.

You get a GV and during it's lifespan the trans, wipers, and ac all blow.
Each time they blow you recall "Oh yeah, I remember CR telling me that
this was going to blow" After the 3rd thing you are thinking "Damn, CR
is accurate" You completely forget they also predicted the floormats
were going to blow, and they didn't.

I could prepare a list of the top ten things that will blow on any
car you buy, and chances are that 3 items on that list will come
true. That does not make me dead on about those cars.

With CR, you have to look at their trends. CR has always been biased
against
domestics and towards imports. At the same time imports have always been
a minority of new car sales. This has been going on for at least 30 years.
If
domestics really sucked as bad as CR claims, (compared to imports) everyone
would have stopped buying them by now..

Ted

Ted Mittelstaedt wrote:

> The only survey that would have meaning is a satisfaction survey that
> polled the numbers of people happy with their cars in each group, then
> normalized for market share/number of cars sold/geographic distribution
> of vehicles sold/etc/etc/etc. But then, you would have a scientific
> survey, which would have real meaning, but would not allow you to
> make up flashy headlines that sell magazines.

Normalized? Okay, rates would be the same thing?

What flashy headlines? It's just an annual roundup
because no one else does this. It's unfortunate that
we have reviews of football players and restaurants
but not what's really important, cars and doctors!

> If the majority of domestic car owners that are happy with their cars
> see a magazine title in the rack that says "owner surveys say imports
> are pieces of ****" why are they going to buy it, they already know that.
> However, if the title says "owner survey says domestics are POS"
> then the domestic car owner thinks "Damn, that's bull****, what are
> those peckers talking about" and picks up the ragazine to look at
> it. Since theres more domestic car owners than import car owners,
> you have more magazine sales if you wank off the majority, rather than
> the minority.

Most people don't read CR. I rarely see someone pickup CR in a
place like Walmart or supermarket.

> > I think this touches on the potential problem of
> > /heteroscedasticity/homeoscedasticity which means the standard errors
> > are or are not randomly distributed? If the errors were random or
> > homeoscedastistic then the sampling was most likely good, even if not
> > random. Is that so? Do you remember statistical inference? I'm trying
> > to.
> >
>
> All of this is bull**** because it is very easy to get a large group of
> humans to believe the WRONG thing. For example, most Americans
> believed that we were going into Iraq because of WMDs, not because
> of oil. That didn't make it any truer.

Hmmm, you're getting angry? This is not fulminous fustigations you
know.
I'm not trying to be funambulist and I'm not sure about these F-words


In the past, CR was right because American cars were poorly made in
comparison to Japanese cars. But that's changing. And with Japanese
factories on our shores, the difference is narrowing.

> If you are talking about sampling apples or something, then your theorm
> applies, but it goes out the window on a survey of people.

Not so fast, my internet interlocutor. Are you saying that people are
not apples? Really. They both have DNA, just a difference of a few
genes.

Actually, what I wrote is true for all surveys. It's by studing the
dispersion of the errors that you can uncover bias.

Or so the economists tell me. I'm not kidding you. But you won't
believe me so run it by a few statisticians you know, fair enough?

> Most people believed the world was flat one time and a giant study
> based on that would have proven it was flat. Fortunately, even back
> then, enough intelligent people knew this argument was bull**** and
> so discovered the New World.

Well, it took the Catholic Church only 300 years to forgive Galileo
for saying the earth revolved around the sun. Now Columbus knew from
his mother's side that the world was round. Why? She was Jewish and
the Jewish, who were fleeing the Spanish Inquisition, had already
been to China and had heard about a new world in the west. The Spanish
Inquistion started in 1492. Spain went downhill after throwing out
the Moors and the Jewish. Dummies. His mother was from Majorca.

> > Generally most polls, like Harris or Gallup, use about 3,000 subjects
> > to handle the USA. According to stat theory, 1,200 with the normal
> > distribution curve, should be enough. So why the over-sampling? It's to
> > make sure the random gets enough since it's quite possible if you play
> > it close to the bone, you miss something.
>
> That isn't why they oversample. They oversample because they are
> telephoning each subject, and since some people don't answer the phone,
> you aren't guarenteed that the number of people who are NOT responding
> is randomly distributed. Most of the time it isn't, so you have to
> oversample
> so you can toss out some respondents in order to compensate for the
> nonrandom distribution of the nonrespondents.

That's interesting. Are you sure? Just asking. I though the
oversampling was to sample less represented people who should have
been in the sample but that particular sample came up short.

> And you are just wrong, you cannot use the kinds of statistical numbers
> arguments that apply to sampling manufactured goods on an assembly
> line to apply to sampling people.

Why not. What's the difference. You have something against us
non-humans?

> All of those auto experts and magazines make their money by selling
> their results, and nobody buys results that tell them what they already
> know. They buy results that seem different than what Real Life tells them
> because they are curious to understand why it is that a survey tells
> them that what they know from direct exposure is wrong. Thus, there
> is ALWAYS a strong financial motivation among the peddlers of
> these surveys to skew the results to be controversial, so that they
> make more money selling surveys.

Nobody is perfect but when push comes to shove, their findings are
pretty accurate. I have found that to be so. They do make mistakes
but not that much when it came to the cars that I actually owned!

> I see this in business all of the time. About once a quarter in some
> meeting a marketing guy will drag out some expensive survey that
> he wasted the companies money on, claiming that some new, unknown
> product that nobody ever thought that they needed or wanted before,
> is now poised to become the next big seller and we gotta catch the
> wave now. Of course, the survey peddler gussies up the survey with
> colorful graphs and expensive bindings, but it still is nothing more
> than a product designed to sell copies of itself.
>
> The ONLY surveys that have any validity whatsoever are the ones
> that get published in a scientific journal for someone's doctorate
> thesis, or some such, because those people do not make money by
> selling more and more thesis copies, and those surveys are subject
> to exhaustive scientific review and must withstand many challenges.
> They also must make the sources the surveys were compiled from,
> available.
>
> The CR car surveys, like most industry surverys, are subject to NO
> outside peer review. CR doesen't even bother to print survey
> rebuttals, to give the users opposing viewpoints. They are
> entertainment value only.

It's not that easy here. The peer-reviewed surveys would be surveys
that by default are published in peer-reviewed journals. That's by
default. There are no peer-reviewed surveys, per se.

Now Ph.D. surveys are not quite peer-reviewed in the same sense.
They are reviewed by other members of the Ph.D. committee but that's
nowhere as rigorous as a peer-reviewed article.

Gallup and Harris are not bad surveys.

Probably the best survey I like is NORC which is the National Opinion
Research Center done by University of Chicago and published by Roper.
Huge survey, ongoing since 1972 which is neither peer-reviewed nor
anything, but has the best data. They do a good job. Thousands of
questions and rigorous sampling methods. I like them and use their
data.

> You get a GV and during it's lifespan the trans, wipers, and ac all blow.
> Each time they blow you recall "Oh yeah, I remember CR telling me that
> this was going to blow" After the 3rd thing you are thinking "Damn, CR
> is accurate" You completely forget they also predicted the floormats
> were going to blow, and they didn't.

> I could prepare a list of the top ten things that will blow on any
> car you buy, and chances are that 3 items on that list will come
> true. That does not make me dead on about those cars.
>
> With CR, you have to look at their trends. CR has always been biased
> against
> domestics and towards imports. At the same time imports have always been
> a minority of new car sales. This has been going on for at least 30 years.
> If
> domestics really sucked as bad as CR claims, (compared to imports) everyone
> would have stopped buying them by now..
>
> Ted

CR is biased against the big cars and the gas guzzling and the sloppy
manufacturing methods. That ain't a bad thing, Ted-dude. If I may call
you that

Ted Mittelstaedt wrote:

> > wrote in message
> oups.com...
>
>
>>That's true but irrelevant, if I may use so harsh a word , because we
>>are describing cars, not the general American public! This ain't no
>>voter survey, it's a car survey.
>>
>
>
> And your ignoring the entire principle of why a self-selected survey is
> bogus.
>
> The people who are driving vehicles with no problems have little incentive
> to return a survey and so toss it. The people with vehicles that have lots
> of
> problems - either due to the vehicle flaw or their own stupidity in
> maintaining
> the vehicle - are mad about it and have an axe to grind and are very happy
> to fill out a survey to "get even with the ****ter who made my POS car"
>
> There's more domestics than imports so if the same pecentages of vehicles
> in each group have problems, then there's more quantity of problem
> domestics than problem imports, so you get more surveys bitching about
> problem domestics.

Yes, not normalizing for the number of vehicles of a given type is just
one error in CR's results. Another is that they survey, at least they
used to, only cars that were five or fewer years old. I've thrown their
surveys in the trash now unopened for at least a dozen years now so
maybe they have widened the age window.

My one and only Japanese vehicle was an 84 Accord purchased new. It was
quite reliable for the first 5 years, and my CR surveys reflected that,
but then it literally self-destructed, including the engine, shortly
after that. However, when I got my next CR survey, my car was now too
old to be included so I never got to report about the thing falling
apart at 6 years and less than 90,000 miles! That was when I started
pitching their annual survey without even opening it.

I've found my American cars to often be a little more trouble during the
first 5 years as compared to the Accord, but the longer-term reliability
and durability have been much better than the Accord. The Accord is the
only vehicle that I've owned for fewer than 100,000 miles since I
started buying cars 30 years ago. I traded it at 90K for an 89 Acclaim
that was a far superior car even though it wasn't rated highly by CR.


Matt

Matt Whiting wrote:
> Yes, not normalizing for the number of vehicles of a given type is just
> one error in CR's results. Another is that they survey, at least they
> used to, only cars that were five or fewer years old. I've thrown their
> surveys in the trash now unopened for at least a dozen years now so
> maybe they have widened the age window.

Okay, I gather rates is the same thing. They don't included how many
vehicles and how many questionnaires. Something I tried to look up. A
rate is a normalization, a sort of index, that could be applied across
the board. Just the raw number of responses and then a rate would have
been nice. Like this vehicle has 200 responses, not sure what the
normalization would offer just a simple raw reporting of N = or how
many responses were obtained for that particular vehicle as opposed to
the whole. I guess normalization would give a neater 1 to 100 range.

> My one and only Japanese vehicle was an 84 Accord purchased new. It was
> quite reliable for the first 5 years, and my CR surveys reflected that,
> but then it literally self-destructed, including the engine, shortly
> after that. However, when I got my next CR survey, my car was now too
> old to be included so I never got to report about the thing falling
> apart at 6 years and less than 90,000 miles! That was when I started
> pitching their annual survey without even opening it.
>
> I've found my American cars to often be a little more trouble during the
> first 5 years as compared to the Accord, but the longer-term reliability
> and durability have been much better than the Accord. The Accord is the
> only vehicle that I've owned for fewer than 100,000 miles since I
> started buying cars 30 years ago. I traded it at 90K for an 89 Acclaim
> that was a far superior car even though it wasn't rated highly by CR.
>
>
> Matt

That's been my experience in a way. My Toyota would not start. Might
have been something as trivial as a coil. Could not get the coil off
since it was kind of wedded to the distributor or something, so I
junked it at 140,000 miles. But I bought it around 90,000 miles? so not
a bad thing. I'm sorry I junked it because it could get 37 mpg! And a
very reliable car, but not too comfy.

However I have bought Chrysler cars new and old, and I think they about
all went farther, 200,000 and 300,000 miles but some very early design
defects used to drive me nuts. In the 70's and early 80's I would
almost state that there was electrolysis in the car to rust out the
back quarter panels. But that's cosmetic although annoying. The
electrical systems were not too good. The transmissions back there were
like tanks with the six cylinder engines, which were indestructible. I
could not hurt my slant six engines. 2 quarts of oil low, no problem!
Hear that loud knock? Just throw in a couple of quarts of oil.