home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!think.com!rpi!crdgw1!rdsunx.crd.ge.com!ilium!molnar
- From: molnar@ilium.crd.ge.com (Karl J Molnar)
- Newsgroups: rec.autos
- Subject: Re: Before Buying Japanese.
- Message-ID: <1993Jan23.210225.2725@crd.ge.com>
- Date: 23 Jan 93 21:02:25 GMT
- References: <C17oxK.FKF@ccu.umanitoba.ca> <1993Jan21.194635.7009@kocrsv01.delcoelect.com> <1993Jan22.142044.20423@magnus.acs.ohio-state.edu> <1993Jan22.220535.10804@leland.Stanford.EDU>
- Sender: usenet@crd.ge.com (Required for NNTP)
- Reply-To: molnar@crd.ge.com
- Distribution: na
- Organization: GE Corporate R&D Center, Schenectady, NY
- Lines: 53
- Nntp-Posting-Host: ilium.crd.ge.com
-
- In article <1993Jan22.220535.10804@leland.Stanford.EDU>, (Theodore Chen)
- tedebear@leland.Stanford.EDU writes
-
- >
- >In article <1993Jan22.142044.20423@magnus.acs.ohio-state.edu>
- jnielsen@magnus.acs.ohio-state.edu >(John F Nielsen) writes:
- >>In article <1993Jan21.194635.7009@kocrsv01.delcoelect.com>
- c23st@kocrsv01.delcoelect.com (Spiros >Triantafyllopoulos) writes:
- >
- >>>
- >>>'CR Statistics' is an oxymoron.
- >>>
- >>>Unless CR can prove that they have used a sample population
- >>>representative of the actual population of course...
- >>
- >>Why is that necessary for what they are trying to show. Unless the
- >>people answering the questionaires are morons or like to lie, the
- >>data is useful, it is not that hard to know when you got a repair.
- >>
- >i was going to stay out of this one, but this seems painfully obvious.
- >CR is trying to rate the cars relative to each other in a number of
- >areas, by using data obtained from questionnaires. you need a
- >certain sample size in order to be mostly certain that your results
- >are representative of the population as a whole. to use an extreme
- >example, i'm sure you would reject a rating based on a single
- >survey response. how about two? how about three?
- >as you increase the sample size the probability of incorrect results
- >decreases (which means that the probability of correct results increases).
- >at some arbitrary level we say we're satisfied and use that as our
- >minimum sample size. in practice, people more than the minimum sample
- >size because they usually have to throw some of the responses away
- >(invalid responses).
- >
- >the point is, even if the people are perfectly truthful and unbiased
- >in their response (and i am dubious about this), random chance may
- >cause their results to be wildly different from the results of the
- >whole group. this is a critical issue in the reliability of CR's
- >ratings. if the sample size is too small, it's not going to be much
- >more reliable than asking one or two people. the information they
- >give may still be useful, but it will be lacking in authoritativeness.
- >
- >-teddy
- >
-
- Certainly the sample size is important, but I think the point is that
- since the sample population may not be representative of the entire
- population, the results may be skewed. For example, if they only surveyed
- people from southern California, then the body rust-through results
- certainly wouldn't mean a whole lot. Does CR publish their criteria
- somewhere as to how they go about getting their results?
-
- Karl Molnar
- molnar@crd.ge.com
-