Custom Tables and Multiple Response Sets: Base issues

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

toscho
Posts: 2
Joined: Tue Sep 08, 2015 10:13 am

Custom Tables and Multiple Response Sets: Base issues

Postby toscho » Tue Sep 08, 2015 10:35 am

Hi there,

I've encountered the following issue. As I don't know the solution, it was quite difficult to choose the right board, so please move this post if necessary.

Background
I have survey answers with filtered questions for awareness and usage (let's say "select all brands you've ever heard of" --> (based on the selection) "which of the services have you ever used")
As a result I have (let's say 10 there are 10 services in total) 10 Vars for Awareness and 10 Vars for Usage (1 Var for each brand)
First Var is coded as [0=No 1=Yes]
Second Var is coded as [0=No 1=Yes Missing] whereby all cells are "missing" if "No" has been chosen in Var 1
Now I wanted to create a custom table showing me the % for all services used (so basically Var2) based on all panelists having answered the question (for this brand).

To make it with an example: 100 people were asked in total, 50 stated to know a service, 25 stated to have used it.
What I want to have in the custom table for Var2 is now n=25 | "50%" (as 50% of those 50 who were asked)

What I did
I created a multiple response set (MRS) for Var2 with default settings
I created a custom table with the MRS as rows and several breakouts (let's say countries, age-groups, total) as columns

What I get
The output is a table with correct n (25 in the example), but wrong % (25% in the example instead of 50%)
Although it is stated for custom tables that missings are always ignored, they are obviously included in the base and I don't know how to solve that (already tried EMPTY = EXCLUDE in the Syntax)

What I did to solve this issue
This is a Syntax:

Code: Select all

CTABLES
  /VLABELS VARIABLES=$B3c CNT Rec_CP_yesno Rec_cp_age_split Rec_mum_yesno DISPLAY=LABEL
  /TABLE $B3c [C] BY CNT [C] > (Rec_CP_yesno [C][COUNT F40.0, COLPCT.COUNT PCT40.1] + 
    Rec_cp_age_split [C][COUNT F40.0, COLPCT.COUNT PCT40.1] + Rec_mum_yesno [C][COUNT F40.0, 
    COLPCT.COUNT PCT40.1]) + CNT [C][COUNT F40.0, COLPCT.COUNT PCT40.1] + Rec_CP_yesno [C][COUNT F40.0, 
    COLPCT.COUNT PCT40.1] + Rec_cp_age_split [C][COUNT F40.0, COLPCT.COUNT PCT40.1] + Rec_mum_yesno 
    [C][COUNT F40.0, COLPCT.COUNT PCT40.1]
  /CATEGORIES VARIABLES=$B3c  EMPTY=EXCLUDE
  /CATEGORIES VARIABLES=CNT Rec_CP_yesno Rec_cp_age_split Rec_mum_yesno ORDER=A KEY=VALUE 
    EMPTY=EXCLUDE.
Whether Empty is included or not doesn't change the Column %

Curiously enough: When I don't use a MRS but simply put all variables in the custom tables´in the row fields, data are correct(but the output is awful as it also shows all "No" answer %, etc.)
I tried now all different settings for MRS and custom tables I could find, but I never get the right % results within the custom tables when using MRS as input

What might cause the issue?
I think I'm either missing a setting in the custom table to make sure bases are computed correctly or I create the MRS in a wrong way.

Any ideas highly appreciated.

---- EDIT----
To make it more tangible, I upload a very simple dataset (3 vars for a MRS, one country var and a total var as breakout vars)
Wit the following Syntax I get a Custom table:

Code: Select all

CTABLES
  /VLABELS VARIABLES=$test DISPLAY=LABEL
  /TABLE $test [C][COUNT F40.0, COLPCT.COUNT PCT40.1]
  /CATEGORIES VARIABLES=$test  EMPTY=INCLUDE TOTAL=YES POSITION=AFTER.
  
which yields the attached result with n=9 (where it should be rowwise n1,2,3 = 8,7,6) and the corresponding wrong %, while the counts of 1 are correct.
I played a little bit around and have absolutely no idea on which basis the sample size "9" is defined... Might be sth like "at least 2 ticks of the 3 vars", but that doesn't make any sense to me.



Best,
toscho
You do not have the required permissions to view the files attached to this post.
JonPedersen
Posts: 117
Joined: Wed May 25, 2011 7:07 am
Contact:

Re: Custom Tables and Multiple Response Sets: Base issues

Postby JonPedersen » Mon Sep 14, 2015 10:22 am

Hi,
The 9 in your sample size comes from line 5 in your data file, namely the one with no occurence in any variables. In order to get 10 you will have to make a variable "No answer" or somesuch that is 1 whenever the other three are all 0 or missing.
Best
Jon
toscho
Posts: 2
Joined: Tue Sep 08, 2015 10:13 am

Re: Custom Tables and Multiple Response Sets: Base issues

Postby toscho » Tue Sep 15, 2015 11:15 am

Hi Jon,

ok, that's a good start to at least get % of the total sample by adding a "total=1" Var to each MRS.

You probably don't have an idea why I can't obtain % based on non-Nans per Variable instead of the whole set level only?


Best,
toscho
JonPedersen
Posts: 117
Joined: Wed May 25, 2011 7:07 am
Contact:

Re: Custom Tables and Multiple Response Sets: Base issues

Postby JonPedersen » Thu Sep 17, 2015 6:33 am

Hi,
I am not sure if you can wit multiple response. I think the multiple response presumes that you have a common base. In most cases you have although there are some where you don't.
It should be possible to do with an ordinary table, unless you want to base percentages on total answers.
jon

Who is online

Users browsing this forum: No registered users and 1 guest

cron