multiple response analysis how to report %

[Spice Weasel]

I have a data set that contains a number of multiple response items, such as:

Which of the following professionals are on staff in your agency?(Tick
all that apply.)
-addiction counsellor
-physician (non-psychiatrist)
-psychiatrist
-psychologist
-social worker
-other health care professional
-other professional.

I have been asked to provide a rationale for whether the % reported
for each option should be the % of responses or the % of respondents.
A review of my stats texts and a quick google search have not yielded
any information on the strengths and weaknesses of either approach.

I'd be very grateful if anyone can point me to a discussion on this.

Thanks.

Pat

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Art Kendall]

I do not know offhand of a source on this.

It all depends on what questions you are asking.
If you are interested in what percent of the staff members are  in the
various occupations then you would use responses as the base.
If you are interested in what percent of the agencies have staff in the
various occupations then you would use cases as the base.

If are interested in both perspectives you would run the tabulation both
ways.

The same type of logic applies if you are crossing a multiple response
variable (e.g., occupation) by an ordinary variable (e.g., city) or
another multiple response variable (e.g., sources of clients, or
presenting problems).

Art Kendall
Social Research Consultants

On 11/26/2010 9:12 AM, Pat C wrote:

> I have a data set that contains a number of multiple response items, such as:
>
> Which of the following professionals are on staff in your agency?(Tick
> all that apply.)
> -addiction counsellor
> -physician (non-psychiatrist)
> -psychiatrist
> -psychologist
> -social worker
> -other health care professional
> -other professional.
>
> I have been asked to provide a rationale for whether the % reported
> for each option should be the % of responses or the % of respondents.
> A review of my stats texts and a quick google search have not yielded
> any information on the strengths and weaknesses of either approach.
>
> I'd be very grateful if anyone can point me to a discussion on this.
>
> Thanks.
>
> Pat
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Spice Weasel]

Thanks, Art. That's helped to clarify my thinking.

On Fri, Nov 26, 2010 at 9:44 AM, Art Kendall <[hidden email]> wrote:

> I do not know offhand of a source on this.
>
> It all depends on what questions you are asking.
> If you are interested in what percent of the staff members are  in the
> various occupations then you would use responses as the base.
> If you are interested in what percent of the agencies have staff in the
> various occupations then you would use cases as the base.
>
> If are interested in both perspectives you would run the tabulation both
> ways.
>
> The same type of logic applies if you are crossing a multiple response
> variable (e.g., occupation) by an ordinary variable (e.g., city) or another
> multiple response variable (e.g., sources of clients, or presenting
> problems).
>
> Art Kendall
> Social Research Consultants
>
> On 11/26/2010 9:12 AM, Pat C wrote:
>>
>> I have a data set that contains a number of multiple response items, such
>> as:
>>
>> Which of the following professionals are on staff in your agency?(Tick
>> all that apply.)
>> -addiction counsellor
>> -physician (non-psychiatrist)
>> -psychiatrist
>> -psychologist
>> -social worker
>> -other health care professional
>> -other professional.
>>
>> I have been asked to provide a rationale for whether the % reported
>> for each option should be the % of responses or the % of respondents.
>> A review of my stats texts and a quick google search have not yielded
>> any information on the strengths and weaknesses of either approach.
>>
>> I'd be very grateful if anyone can point me to a discussion on this.
>>
>> Thanks.
>>
>> Pat
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>> [hidden email] (not to SPSSX-L), with no body text except the
>> command. To leave the list, send the command
>> SIGNOFF SPSSX-L
>> For a list of commands to manage subscriptions, send the command
>> INFO REFCARD
>>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[John F Hall]

Pat

I'm not sure if this will help, but section 3.3 of the (syntax-based) SPSS
tutorials on my website is on multiple response.  (See:
http://surveyresearch.weebly.com/33-multiple-response.html )

The intro 3.3.1 is well worth reading as background, and will possibly
address the questions you have been asked about rationale.

3.3.2 lists examples of MR questions used in the 1986 British Social
Attitudes survey and the  different coding schemes used to create the raw
data file.

3.3.3 is a fully worked example, with full colour screenshots at each step,
using SPSS 15 to analyse MR questions from a small survey of 15-16 year-old
pupils in a secondary school.  One example uses MR in dichotomous mode, to
summarise, in a single table, several items from a Likert-type scale of
attitudes to women.

The third slide show accompanying my 2006 presentation "Old Dog, Old Tricks"
to ASSESS (European SPSS users) contains a critique of SPSS usage in the
2002 European Social Survey with examples of MR for questions on personal
experience of discrimination.

I'm now using 18 so some of the examples may need updating.  I'm currently
working on an exercise for the 1986 BSA survey and will then add some
homework on the same or similar questions in the 1989 survey.   OK, so the
data are more than 20 years old, but the logic and analysis are what count.

I used the 1986 and 1989 surveys on the postgrad Survey Analysis Workshop I
designed and taught from 1976 until I (early) retired in 1992.  I'm busy
converting and updating a colossal ampount of course materials from
WordStar4 to Word and from SPSS-X 4 on a Vax mainframe to SPSS for Windows
on a PC, but as soon as I've got most of this done, I want to start using
later waves of the BSA and also data from the European Social Survey and the
GSS.

All materials on the site are available for free download.  I can't send
attachements via the list, so may send you (and anyone else) some of my
draft MR material separately.

John Hall
[hidden email]
http://surveyresearch.weebly.com


----- Original Message -----
From: Pat C
To: [hidden email]
Sent: Friday, November 26, 2010 3:12 PM
Subject: multiple response analysis how to report %



I have a data set that contains a number of multiple response items, such
as:

Which of the following professionals are on staff in your agency?(Tick
all that apply.)
-addiction counsellor
-physician (non-psychiatrist)
-psychiatrist
-psychologist
-social worker
-other health care professional
-other professional.

I have been asked to provide a rationale for whether the % reported
for each option should be the % of responses or the % of respondents.
A review of my stats texts and a quick google search have not yielded
any information on the strengths and weaknesses of either approach.

I'd be very grateful if anyone can point me to a discussion on this.

Thanks.

Pat

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Allan Reese (Cefas)]

I can't point to a reference or discussion but my own experience is that
reporting the proportion of responses is less likely to be useful but
reporting the proportion of respondents must be done with clarity.

Case 1: "tick all that apply", in which case respondents report a
variable number and the total of responses is a random number.  It might
seem you could report "X% of staff were described as type A", but the
unit of reporting appears to be the institution, and my first
inclination goes with "institutions reported having between -min- and
-max- professionals on their staff (mean -mean-). Y% of respondents
reported they had staff type A."

Case 2: "list your top three preferences".  Number of responses is
(except for missing) 3 x number of respondents.  I might try, "X% of
respondents reported A as their first preference, and Y% had A among
their top three."

The important thing is for the reader (and the author!) to be aware of
what it's a percentage of and whether it's exclusive.  An example came
up last week on the BBC Radio 4 Today programme.  A contributor claimed
"40% of your audience are probably listening on DAB." When this was
analysed, it appeared (this is my memory, check on bbc.co.uk/today
website) to be based on listener surveys that suggested about 33% of
listeners used digital -at some time during the week-.  How the time
spent listening was split between digital and analogue was not reported.

Allan


***********************************************************************************
This email and any attachments are intended for the named recipient only.  Its unauthorised use, distribution, disclosure, storage or copying is not permitted.  If you have received it in error, please destroy all copies and notify the sender.  In messages of a non-business nature, the views and opinions expressed are the author's own and do not necessarily reflect those of the organisation from which it is sent.  All emails may be subject to monitoring.
***********************************************************************************

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ling Ting]

I will be out of office till Nov. 30 with limited access to email.   If you need immediate assistance please contact 479-575-2905.  Thank you.

[E. Bernardo]

Dear all,

 

The degree of closeness between dyad (younger-older pair of siblings) was measured using a five-point Likert scale response format.  A dyad can be classified as either childhood (n=20) or adolescence (n=20). That is, a dyad can belong only to one classification (for example, both the younger and older are adolescence).   I am using a repeated measures GLM such that the dyad(younger-older pair of siblings) is considered as within-subjects factor whle the level (childhood and adolecence) is considered as the between-subjects factor.  Three concerns:

(1) Is repeated measures GLM appropriate for my variables?

(2) I am sure my sample size is inadequate, please suggest an alternative model (if any).

(3) What is the APA format for presenting GLM results?

 

Thank you in advance for your help.

[Maguin, Eugene]

Eins,

On 3): Basically, you have to describe your analysis model in sufficient
detail that someone else could replicate your work, describe any relevant
assumptions checking, report the results. Beyond that this is too
complicated to go into in a reply. Look at some articles that used repeated
measures and copy the format.

On 2): Not be coy, but maybe or maybe not. Given a legitimate expectation of
large enough effects, your sample would be perfectly adequate. Different
analysis models are possible. I don't judge myself knowledgable enough to
rank them on power.

On 1): A between-within repeated measures (RM) is a natural choice. But, it
may tell you more than you want to know. Your basic RM will give you two
main effects and an interaction. I'd guess that you are really interested in
the between main effect. I'm not quite sure how your measure works but I'd
guess that you are claiming that childhood dyads are 'closer' than
adolescent dyads (or vice-versa) where 'closer' means the difference between
person 1 and person 2 is smaller in the childhood dyads than in the
adolescent dyads. If so, then the within subjects main effect and the
interaction are irrelevant. So, a simpler analysis would be a t-test on the
difference between person 1 and person 2. By the way, you might want to
think about whether you should analyze the difference or the absolute value
of the difference in a t-test.

If differences are there, you want to give yourself the best chance of
finding them. That's a question of the power of different analysis models
given an effect size, N, and alpha. While a t-test is simpler, I can't
confidently say that it is more powerful. Others on this list can; perhaps
they will.

Gene Maguin

________________________________

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Eins Bernardo
Sent: Monday, November 29, 2010 8:00 AM
To: [hidden email]
Subject: Repeated Measures GLM


Dear all,

The degree of closeness between dyad (younger-older pair of siblings) was
measured using a five-point Likert scale response format.  A dyad can be
classified as either childhood (n=20) or adolescence (n=20). That is, a dyad
can belong only to one classification (for example, both the younger and
older are adolescence).   I am using a repeated measures GLM such that the
dyad(younger-older pair of siblings) is considered as within-subjects factor
whle the level (childhood and adolecence) is considered as the
between-subjects factor.  Three concerns:
(1) Is repeated measures GLM appropriate for my variables?
(2) I am sure my sample size is inadequate, please suggest an alternative
model (if any).
(3) What is the APA format for presenting GLM results?

Thank you in advance for your help.


=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ryan]

Eins,

You need to provide more detail around the design, dependent variable,
and the research question(s) of interest. I started to form a response
and realized that there were too many holes to continue. For instance,
how exactly was degree of closeness measured?

Ryan

On Mon, Nov 29, 2010 at 7:59 AM, Eins Bernardo
<[hidden email]> wrote:
>
> Dear all,
>
> The degree of closeness between dyad (younger-older pair of siblings) was measured using a five-point Likert scale response format.  A dyad can be classified as either childhood (n=20) or adolescence (n=20). That is, a dyad can belong only to one classification (for example, both the younger and older are adolescence).   I am using a repeated measures GLM such that the dyad(younger-older pair of siblings) is considered as within-subjects factor whle the level (childhood and adolecence) is considered as the between-subjects factor.  Three concerns:
> (1) Is repeated measures GLM appropriate for my variables?
> (2) I am sure my sample size is inadequate, please suggest an alternative model (if any).
> (3) What is the APA format for presenting GLM results?
>
> Thank you in advance for your help.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD