Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics

Hi. Can I please see your CTABLES syntax, so I can replicate your table exactly? I'm sure I could figure it out from the data file but it would help. Thank you in advance.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue March 22, 2022 12:35 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics

Original Message:
Sent: 3/22/2022 2:14:00 PM
From: Rick Marcantonio
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Hi. Can I please see your CTABLES syntax, so I can replicate your table exactly? I'm sure I could figure it out from the data file but it would help. Thank you in advance.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue March 22, 2022 12:35 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics

Original Message:
Sent: 3/22/2022 2:59:00 PM
From: Jon Peck
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Adding to Rick's post, the video was too fuzzy to make out what the Tables code was.

But from the description, it seem to me that you would do a column proportions or column means test rather than the chi-squared test you apparently used.

--

Jon K Peck
jkpeck@gmail.com

Original Message:
Sent: 3/22/2022 2:14:00 PM
From: Rick Marcantonio
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Hi. Can I please see your CTABLES syntax, so I can replicate your table exactly? I'm sure I could figure it out from the data file but it would help. Thank you in advance.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue March 22, 2022 12:35 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics

You stole my answer. :) It has to do with the mult-response variable.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue March 22, 2022 03:21 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Thank you for the support. I've found the answer:

Since I'm working with "multiple response sets"... then the (C-1)(L-1) rule for degrees of freedom doesn't apply. This rule assumes that the categories are fixed, which is not the case for multiple sets. If the multiple sets are in a row, like my table, and the sort variable is in a column, the degrees of freedom are L*(C-1). In your case 15*(3-1). The formula for Qui2 is not the usual one, the denominator is different (Eij*(1- cj/N)...

Original Message:
Sent: 3/22/2022 2:59:00 PM
From: Jon Peck
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Adding to Rick's post, the video was too fuzzy to make out what the Tables code was.

But from the description, it seem to me that you would do a column proportions or column means test rather than the chi-squared test you apparently used.

--

Jon K Peck
jkpeck@gmail.com

Original Message:
Sent: 3/22/2022 2:14:00 PM
From: Rick Marcantonio
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Hi. Can I please see your CTABLES syntax, so I can replicate your table exactly? I'm sure I could figure it out from the data file but it would help. Thank you in advance.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Tue March 22, 2022 12:35 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics

Original Message:
Sent: 3/22/2022 3:39:00 PM
From: Rick Marcantonio
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

You stole my answer. :) It has to do with the mult-response variable.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue March 22, 2022 03:21 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Thank you for the support. I've found the answer:

Since I'm working with "multiple response sets"... then the (C-1)(L-1) rule for degrees of freedom doesn't apply. This rule assumes that the categories are fixed, which is not the case for multiple sets. If the multiple sets are in a row, like my table, and the sort variable is in a column, the degrees of freedom are L*(C-1). In your case 15*(3-1). The formula for Qui2 is not the usual one, the denominator is different (Eij*(1- cj/N)...

Original Message:
Sent: 3/22/2022 2:59:00 PM
From: Jon Peck
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Adding to Rick's post, the video was too fuzzy to make out what the Tables code was.

But from the description, it seem to me that you would do a column proportions or column means test rather than the chi-squared test you apparently used.

--

Jon K Peck
jkpeck@gmail.com

Original Message:
Sent: 3/22/2022 2:14:00 PM
From: Rick Marcantonio
Subject: RE: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Hi. Can I please see your CTABLES syntax, so I can replicate your table exactly? I'm sure I could figure it out from the data file but it would help. Thank you in advance.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Tue March 22, 2022 12:35 PM
From: David Magalhaes
Subject: Chi-Square with post hoc for multiple choice (degrees of freedom (df value) value)

Dear all,

I've 3 groups that answered between 3 to 5 options in a multiple choice of 15. I'm trying to analyse if there are significant differences between the groups for each option. I've created a multiple question analysis variable with a "yes" or 1 dichotomy and created a custom table. When all the choices are displayed, the degrees of freedom (df value) are 30, but should be 28 ((3-1)*(15-1)=28). What am I doing wrong? For better comprehension I've created a vídeo, and attached the data file if it is simpler or clearer.

Thank you all in advance for the help!

Regards,
David

------------------------------
David Magalhaes
------------------------------

#SPSSStatistics