NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/rec/autos/32723

Internet Message Format | 1993-01-26 | 3.8 KB

Path: sparky!uunet!paladin.american.edu!howland.reston.ans.net!zaphod.mps.ohio-state.edu!rphroy!kocrsv01!c23st
From: [e-mail] (Spiros Triantafyllopoulos)
Newsgroups: rec.autos
Subject: Re: yet another consumer reports article (Re: Before Buying Japanese.)
Message-ID: <[e-mail]>
Date: 25 Jan 93 18:26:14 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail] (Usenet News Account)
Distribution: na
Organization: Delco Electronics Corp.
Lines: 65

In article <[e-mail]> [e-mail] (John F Nielsen) writes:
>>a. questionnaires are mailed to people in proportion to car ownership
>>   (or market share)
>>b. response rates are similar to mailing rates
>>c. response rates are accurate as far as problem reporting goes
>>d. statistical methods are followed properly
>
>What I am contending is that for what they are trying to show,
>all of this is not necessary.

Arg. And you're a scientist :-). First off, (a) is necessary if
you want to have a sample that accurately describes the population.
The way this is checked is usually distribution tests, fairly easy.
(b) ditto. (see note below). (c) response accuracy is important; I
may have the mother of all electrical troubles and it could be a
battery connector for all I know; how 'car-aware' are we? (d) ensures
that all other factors have been eliminated, and only genuine quality
differences emerge (not damage/wear due to driving skills/climate,
for example).

>IF you have 10,000 people responding about problems with their Ford
>Ranger  and 50,000 about problems with their Ford Taurus, patterns
>will show up. And, both with have groups of people who respond
>because they love or hate their machine.

Sounds like the 'coffee causes heart disease' study. Turned out that
if the folks doing the analysis had bothered to consider OTHER factors
in the analysis, they'd seen that most people who drink 5-10 cups a
day are type-A, smoke heavily, eat lots of donuts (I am NOT making
this up btw, this is a true study.) etc. When these factors were
eliminated (again easy to do if you know how) what emerges may
surprise you.

(note) But the bottom line John is that it is a scientific fact that
people satisfied with their cars will NOT respond at the same rate as
people who are NOT. When was the last time you bragged about your
flawless widget A and when was the last time you bitched about the
flawed widget B? If I recall a study done people with complaints were
something like 11 times more likely to respond than people without
complaints...

There is also the issue of sample size; how many CR readers drive K-cars
vs. Accords? and out of these K-car-CR-readers, how many respond vs
how many Accord-CR-readers? Out of how many? and in what proportions?

It is simply for the same reason that call-in audience polls are NOT
accurate. CR or anyone else has no way of controlling the response
rate, or of mapping the responses back to a correct population.

>>In my case, a couple of things. One was that the lawnmower in question
>>was WAY too HEAVY after grass clippings were collected; its engine
>
>Which mower was this?

A Ho-Ho-Honda :-). The SR214. Flawless machine, manual written in remedial
English, I've had to hack it to make it go fast enough. Very reliable
though. Heavy, and even heavier when loaded with clippings. Oil changes
are not easy, eats lots of gas, and despite an Overhead Valve engine it
is underpowered.

Spiros
--
Spiros Triantafyllopoulos                    [e-mail]
Software Technology, Delco Electronics       [phone]
GM Hughes Electronics, Kokomo, IN 46904      [A Different Kind of Disclaimer]