Missing data in multiple response table

[Frank Furter]

I would like to create a table from a group of multiple dichotomous
variables (coded 1/0) using the Custom Tables procedure. Some of the
subjects have not selected any of the options presented to them , and for
them I have generated an additional variable that is 1 when no option was
selected and 0 otherwise. Ideally, I want this variable to appear in the
multiple response table like a user-defined missing value would appear in a
single response table, e.g.

Option A ... N ... Valid % ... Total %
Option B ... N ... Valid % ... Total %
Option C ... N ... Valid % ... Total %
Total valid ... N ... 100% ... Total %
None of the above ... N ... - ... Total %
Grand total ... N ... 100% ... 100%

Is it possible to do this in a table with multiple responses?

Thanks for your thoughts
Andreas

[Ruben Geert van den Berg]

Perhaps you could use something similar to this http://www.spss-tutorials.com/include-empty-categories-in-output/?

HTH,

Ruben

[Frank Furter]

Ruben,

thanks very much for your response. The tutorial is, however, about a different issue where you want to include categories into your table although they may have zero counts (in CTABLES there is actually an option for this, "/categories empty=include", that does exactly this without the requirement of a workaround).

Maybe I should state my problem more clearly: Imagine that you are doing a survey about the newspapers that people have read regularly during the last year. Some may have read one paper, some more than one, and some none at all. You could then create a dichotomous variable for every newspaper mentioned and code 1 if read and 0 (or missing) if not. Then you generate a table (using the Custom Tables option) based on the multiple dichotomies.

The thing that interests me is what happens to the survey participants who do not read any paper: If you proceed as described above, these participants will not appear in the table at all; the absolute total count will be the number of participants who read at least one paper, and the percent value will also be based on this number in the denominator (this will, e. g., give you the percentage of respondents who have read newspaper "X" among those who have read at least one paper regularly). Alternatively, you could create another variable that is coded 1 for every respondent who has not read any paper regularly. If you include this into your multiple dichotomies, the "non-readers" will appear as a separate category and the percentages will be based on the total number of survey participants (this will, e. g., give you the percentage of respondents who have read newspaper "X" among those who have participated in your survey, regardless of whether or not they have read any paper).

I would now like to create a multiple response table that has both percent columns side by side (because both may be interesting in their own right) instead of having to generate two separate tables for this. In a table with a single response item this would be no problem at all because there are different statistical options for both cases, COLPCT.VALIDN and COLPCT.COUNT, which can be used side by side in one table. I would like to do the same in a multiple response table, which would probably require declaring one of the multiple dichotomy variables to identify 'missing' cases.

Andreas

[Jon K Peck]

I can think of two ways to attach this problem.  First, use the PCOMPUTE subcommand to create a column of adjusted percentages of the inclusive MR set.  I am not positive that PCOMPUTE has enough power to do this, but take a look.

If that does not work, then the STATS TABLE CALC extension command could create a new column (V21 or later) that is a function of the inclusive MR set values.  It would probably require a small custom function plugged in.  If you want to send me an spv file containing the table with all the counts and percentages in it, I can help with that.


Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
phone: [phone]




From:        Andreas Voelp <[hidden email]>
To:        [hidden email],
Date:        07/23/2014 12:42 AM
Subject:        Re: [SPSSX-L] Missing data in multiple response table
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Ruben,

thanks very much for your response. The tutorial is, however, about a
different issue where you want to include categories into your table
although they may have zero counts (in CTABLES there is actually an option
for this, "/categories empty=include", that does exactly this without the
requirement of a workaround).

Maybe I should state my problem more clearly: Imagine that you are doing a
survey about the newspapers that people have read regularly during the last
year. Some may have read one paper, some more than one, and some none at
all. You could then create a dichotomous variable for every newspaper
mentioned and code 1 if read and 0 (or missing) if not. Then you generate a
table (using the Custom Tables option) based on the multiple dichotomies.

The thing that interests me is what happens to the survey participants who
do not read any paper: If you procede as described above, these participants
will not appear in the table at all; the absolute total count will be the
number of participants who read at least one paper, and the percent value
will also be based on this number in the denominator (this will, e. g., give
you the percentage of respondents who have read newspaper "X" among those
who have read at least one paper regularly). Alternatively, you could create
another variable that is coded 1 for every respondent who has not read any
paper regularly. If you include this into your multiple dichotomies, the
"non-readers" will appear as a separate category and the percentages will be
based on the total number of survey participants (this will, e. g., give you
the percentage of respondents who have read newspaper "X" among those who
have participated in your survey, regardless of whether or not they have
read any paper).

I would now like to create a multiple response table that has both percent
columns side by side (because both may be interesting in their own right)
instead of having to generate two separate tables for this. In a table with
a single response item this would be no problem at all because there are
different statistical options for both cases, COLPCT.VALIDN and
COLPCT.COUNT, which can be used side by side in one table. I would like to
do the same in a multiple response table, which would probably require
declaring one of the multiple dichotomy variables to identify 'missing'
cases.

Andreas



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Missing-data-in-multiple-response-table-tp5726786p5726794.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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