Date: Fri, 19 Jul 2002 10:18:22 -0400
Reply-To: "Cohen,Diane" <dc42@EXCHANGE1.DREXEL.EDU>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "Cohen,Diane" <dc42@EXCHANGE1.DREXEL.EDU>
Subject: Re: 80% criterion: survey research
Content-Type: text/plain; charset="iso-8859-1"
Giving a "too personal to answer" option would have resulted in far more unanswered items, I assure you. We did not get "a lot" of unanswered questions, but I am certain we would have had we used your advice. I strongly recommend not offering that option. Odd that you thought we'd not piloted the survey (which of course we did). The one question giving some persons trouble even tho guaranteed anonymity, was "how often in the past six months have you engaged in pleasurable sexual activity" - no big surprise why it wasn't answered. dgc
-----Original Message-----
From: David Lindsay [mailto:spss2002@sdresearch.com]
Sent: Friday, July 19, 2002 9:43 AM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: 80% criterion: survey research
On 19 Jul 2002 at 8:26, Cohen,Diane wrote:
> I've had a similar problem: we distributed a quite lengthy well-being
> survey (137 items plus an open ended question at the end). It contained
> about six subscales, some requesting quite personal information (an
> anxiety scale, a depression scale, a life satisfaction scale, etc.).
> Respondants reacted in a a couple of ways: skipping some subscales
> altogether, or not answering specific questions in a subscale.
With these type of personal questions you are bound to get some
skipping. Perhaps you should have a "too personal to answer
option" so you got better completion of the questionnaire. If
you get a large amount of n/a's it could imply that the qu'aire
should have been piloted; or you re-interview the people to
find out what was the reason for poor completion.
> I
> checked with colleagues on this listserve and elsewhere, and came to the
> conclusion that it was acceptable to use a respondant's answers if they
> answered at least 75% of a subscale's items.
In answer to this point, and the original point, I would ask
you why you think that you have the "right" to exclude
respondents? The fact that they chose not to answer is valid
research information; by taking the decision to exclude people
you will have biased the results slightly.
If you have evidence that the respondent has deliberately
ruined the survey somehow, then that would be a good reason to
exclude them.
--
David Lindsay
Significantly Different! Research, Reading,Berks. GB
Market Research Services & Consultancy
http://www.SDResearch.com/
