"C. E. White" <cewhite3@removemindspring.com> wrote
> "Elle" <honda.lioness@gmail.com> wrote in message
> news:4Ru0l.4052$%z5.3924@newsfe09.iad...
>> "C. E. White" <cewhite3@removemindspring.com> wrote
>>> "Lloyd" <lparker@emory.edu> wrote
>>>> Why don't you READ CR's explanation of their analysis
>>>> before spouting
>>>> off and looking stupid?
>>>
>>> http://www.consumerreports.org/cro/c...ity-faq_ov.htm
>>>
>>> I have read it and I still believe the little circle
>>> over
>>> emphasize small differences, that may not be
>>> statistically
>>> valid. In the discussion they talk about the average
>>> circle representing a failure rate of 2.5% (not an
>>> actual
>>> value, but their example).
>>
>> In what part did you see this mention of 2.5 percent? I
>> see
>> it only in section 4.2, and it says nothing like what you
>> say.
>
> Yes, section 4.2. I expressed myself poorly. In the CR
> example of what happens with small problem rates they use
> the 2.5% average failure rate as an example (which is why
> I said "not an actual value, but their example").

It is not an example. Nor do they call it a "failure rate."
It is a "problem rate."

Section 4.2 does not say anything like what you are saying.

>>> Do you think that a sample size of 100 vehicles is
>>> enough
>>
>> This is the minimum required sample size. If you read the
>> site linked above, it notes that 200-400 cars per model
>> year
>> is usual.
>
> And this is still a very small sample size when you are
> talking about very small actual differences in failure
> rates.

Problem rates. Not when one is trying to establish whether
the differences between two models are statistically
significant.

I think this is a very honest and fair statement by CR:
"While the difference between a [fully colored red circle]
and a [half colored red circle] may be small, a pattern of
several less-than-perfect trouble spots in a brand new car
should be cause for concern and does not bode well for a
model's long-term reliability."

>> I disagree with other points you are making.
>
> Which points are those?

Pretty much all of them, because they are based on the large
errors you made above. IOW, your premise is wrong.

I do not have a problem with anyone calling the samples CR
uses biased, as long as they call nearly all samples used in
any such study biased. CR's survey is what it is. I have not
seen a better survey out there.

Really, CE White, you are entitled to your opinion. I do not
expect you to buy mine. I am responding for others, as an
engineer with publications, advanced degrees, much education
in statistics, and years of engineering experience. Not that
these make me more of an authority. More that others
poo-pooing the CR study are saying they have such
credentials. They're wrong, AFAIC.