I don't really understand your syntax. Your aux_sig value is a constant with value -1..
What I was suggesting is that you create a variable like this.
compute group = x eq 1.
for a group of all the 1 values for x with group declared ordinal or nominal.
Then use that in CTABLES with
/table group by someColumnVariable
That gives you two categories - one the particular group of interest and the other all the others.
Then the column proportion test is comparing that group to everything else.
------------------------------
Jon Peck
------------------------------
Original Message:
Sent: Wed July 28, 2021 08:18 AM
From: Konrad Gałuszko
Subject: CTABLES significance test - comparison to grand total
I don't understand to be honest. What should I do to compute categories using your method?
I tried another way (for single answer variables). I created an auxillary variable with an unique value (and label) and also added value label for my variable. Something like this:
define AuxVarSig ()
compute aux_sig = $sysmis.
compute aux_sig = -1.
variable labels aux_sig "Significance compared to total".
value labels aux_sig
-1 " Significance compared to total".
execute.
variable level aux_sig (nominal).
formats aux_sig (f1.0).
add value labels all
-1 "TOTAL".
!enddefine.
AuxVarSig.
mrsets
/mcgroup name = $SINGLE_AUX_MRSET label = "Test"
variables = aux_sig single_answer_variable.
output close all.
ctables
/mrsets countduplicates = no
/table CROSS_VAR [colpct.count pct40.0, totals[count f40.0, colpct.count pct40.0]] by $SINGLE_AUX_MRSET
/slabels position = row visible = no
/categories variables = CROSS_VAR empty = exclude missing = exclude total = yes position = after label = 'TOTAL'
/comparetest type = prop alpha = 0.05 adjust = bonferroni includemrsets = yes categories = allvisible .
The problem is I didn't get the same significance results as before (meaning I hadn't added aux_sig variable and mrset so the comparison was only categorical pairwise).
Original Message:
Sent: 7/26/2021 1:05:00 PM
From: Jon Peck
Subject: RE: CTABLES significance test - comparison to grand total
You have the same issue with that. You could compute two categories - cat1 and not cat1, say, and do a test on that
--
Original Message:
Sent: 7/26/2021 1:03:00 PM
From: Konrad Gałuszko
Subject: RE: CTABLES significance test - comparison to grand total
Okay, I see. What about suppressing comparing categories and, instead, turn on comparing to total only?
Will total compare be available in further versions of SPSS?
Original Message:
Sent: 7/26/2021 11:39:00 AM
From: Jon Peck
Subject: RE: CTABLES significance test - comparison to grand total
The trouble with using the grand total is that it includes the cases in the components you want to test. That's similar, also, in the one-way ANOVA where you can do post hoc tests of each factor level against each other one but not test a factor level against the overall effect.
------------------------------
Jon Peck
Original Message:
Sent: Mon July 26, 2021 09:34 AM
From: Konrad Gałuszko
Subject: CTABLES significance test - comparison to grand total
Hello.
The CSR states that significance testing is only pairwise - meaning there's no possibility to compare column result to grand total (not subtotal). Is the any way to do so after all? Any workaround or extension?
------------------------------
Konrad Gałuszko
------------------------------
#SPSSStatistics