Barry
Actually it can get more complex than the number of options. And, you can
get a range of opinions on the number of options.
There are no simple answers.
Certainly 7 options potentially gives you more variability in the response
set.
Also, folks will argue that having an even number of options avoids the
'mean response' problem.
However, technically these scales are not mathematically continuous, they
are actually categorical
although most folks treat them as continuous.
I agree with Paul, it also depends on the purpose of the survey and the
intended audience.
While it is likely that over a large population responses will regress to
the mean (whatever it is) there are some interesting
response patterns to surveys.
Those with more extreme responses (greater like-dislike, greater love-hate,
etc.) tend to be early responders. These are
the folks that have the strongest opinions and have already formed an
opinion on the topic.
If this is a one time survey (once out with follow-ups) you will likely get
about 60% of the final response rate
in the first 7-10 days. So if you send out 1,000 surveys and get 60 back,
you are likely to get a total of 100 at the end.
Surveys with low response are targeted to the wrong group and/or don't
address questions of importance to the target group.
The format of each item is critical depending on the purpose of the survey.
Do you want to really know what
you are doing well and what needs to be improved? If so, the items should
help you discriminate.
I would suggest a pilot test or two. Try different numbers of options and
different item wordings to see
which pattern works best for answering your research question.
Patient satisfaction is a tough area to truly assess. Patients, even in
anonymous surveys, don't want to
'offend' their physicians and may temper their responses. When I do
patient surveys I always pilot test
the surveys with individuals from the intended target audience and use an
observed response process where I sit with
the respondent and then after each of their answers probe with them what
the understood the item
to mean and then why they answered the way that they did. Also, I am
interested in what other
thoughts they have that are triggered by the item. Sometimes respondents
will say something to the
effect "You asked about A but what is really important to me is B." That
helps drive the final item
formation and selection.
All the best with your survey.
Bill
>>> Paul Alexander <[log in to unmask]> 3/12/2009 7:44 AM >>>
I tend agree with Anne, that with 4 choices, it forces a decision. Maybe 6
choices would add some more nuance and valuable information for analysis.
But yes, people tend to take the middle ground if given the choice when in a
5 point scale. This is a very interesting question and the debate can make
for interesting feedback. Depends on the survey objectives too.
Best,
Paul
--- On Thu, 3/12/09, Anne Gill <[log in to unmask]> wrote:
From: Anne Gill <[log in to unmask]>
Subject: Re: Survey 1.0
To: "Paul Alexander" <[log in to unmask]>
Received: Thursday, March 12, 2009, 11:35 AM
Hello All I am not a statistician but I understood if you have an odd
number in a Likert scale e.g. 5 people will tend to favour the middle number
as the neutral option whereas if you have 4 or 6 they are more likely to
make a decision either way which could be important in some instances.
Regards
Anne
On Thu, Mar 12, 2009 at 11:28 AM, Paul Alexander <[log in to unmask]>
wrote:
Yes, the typical number is 5 with a graded response, the middle most often
'3' on a 1-5 scale.
Best,
Paul
--- On Thu, 3/12/09, Netuveli, Gopal <g.netuveli@IMPERIAL..AC.UK> wrote:
From: Netuveli, Gopal <[log in to unmask]>
Subject: Re: Survey 1.0
To: [log in to unmask]
Received: Thursday, March 12, 2009, 6:49 AM
Majority of the Likert scale items in satisfaction surveys tend to be five.
Gopal Netuveli
PCSM, Imperial Colege London
From: Evidence based health (EBH) [[log in to unmask]] On
Behalf Of Barry Diner [[log in to unmask]]
Sent: 12 March 2009 01:57
To: [log in to unmask]
Subject: Survey 1.0
Sorry I did not give enough information
The survey is patient satisfaction survey looking at specific issue related
to the unique patient population at the Texas Heart Institute.
We want to look at overall satisfaction, expectations and areas of weakness
It will be disseminated online through email
When I asked about scale I meant the number of possible responses 5 or 7 or
more
Thanks
Barry
Yahoo! Canada Toolbar : Search from anywhere on the web and bookmark your
favourite sites. Download it now!
--
Anne Gill, Senior Lecturer,
University of Greenwich
School of Health and Social Care
Room MS218, Southwood Site
Avery Hill Rd
Eltham
London SE9 2UG
Tel: +44 (0)20 8 331 7502
Fax: +44 (0)20 8 331 9941
email: [log in to unmask]
If you have a Freedom of Information request, please direct it to:
[log in to unmask]
University of Greenwich, a charity and company limited by guarantee,
registered in England (reg no. 986729). Registered Office: Old Royal
Naval College, Park Row, Greenwich SE10 9LS.
__________________________________________________________________
Yahoo! Canada Toolbar: Search from anywhere on the web, and bookmark your
favourite sites. Download it now at
http://ca.toolbar.yahoo.com.
|