Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:


------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole
------------------------------

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?


------------------------------
Courtney Cole
------------------------------

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

I think I might leave Semester and recode instructor to 0-1 (Instructor=1 or not).

Did you have any warning messages when you ran these analyses?

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

Original Message:
Sent: Mon February 14, 2022 05:49 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

Hi Rick! Thanks again for responding!
I did not find any error when I ran these the first time. However, I did take your suggestion for Instructor and re-ran it (no more issues with instructor):


------------------------------
Courtney Cole
------------------------------

Original Message:
Sent: Mon February 14, 2022 06:23 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

I think I might leave Semester and recode instructor to 0-1 (Instructor=1 or not).

Did you have any warning messages when you ran these analyses?

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

Original Message:
Sent: Mon February 14, 2022 05:49 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

I couldn't help but try running the data with the condensed categories for semester, and it appears to be working better:


------------------------------
Courtney Cole
------------------------------

Original Message:
Sent: Mon February 14, 2022 07:31 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thanks again for responding!
I did not find any error when I ran these the first time. However, I did take your suggestion for Instructor and re-ran it (no more issues with instructor):

However, Semester still seems to be problematic. My next thought is to combine summer semesters together and combine spring and fall together (summer was condensed in comparison to the regular school year). I think it might just be a matter of too many "dummy variables" going on for this kind of analysis. Would you say that may be a fair assumption?

Thank you!

Courtney

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 06:23 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

I think I might leave Semester and recode instructor to 0-1 (Instructor=1 or not).

Did you have any warning messages when you ran these analyses?

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

Original Message:
Sent: Mon February 14, 2022 05:49 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

My preference is to begin with simpler models and move to more complex only if needed. I don't know your hypothesis but as I see it here, simpler is better.

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

Original Message:
Sent: Mon February 14, 2022 08:04 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

I couldn't help but try running the data with the condensed categories for semester, and it appears to be working better:

It looks like there was just too much going on in my model from before...

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 07:31 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thanks again for responding!
I did not find any error when I ran these the first time. However, I did take your suggestion for Instructor and re-ran it (no more issues with instructor):

However, Semester still seems to be problematic. My next thought is to combine summer semesters together and combine spring and fall together (summer was condensed in comparison to the regular school year). I think it might just be a matter of too many "dummy variables" going on for this kind of analysis. Would you say that may be a fair assumption?

Thank you!

Courtney

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 06:23 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

I think I might leave Semester and recode instructor to 0-1 (Instructor=1 or not).

Did you have any warning messages when you ran these analyses?

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

Original Message:
Sent: Mon February 14, 2022 05:49 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics

Hello again Rick!
I agree--I prefer simpler in this case. Unfortunately, prior literature in my field points to confounding effects with instructor, term (summer vs. regular school year), and design task...so that is my reason for including it (I've been criticized in past paper reviews when these things were omitted, so I have been trying to keep track of them). I did run my analyses with less categories, though, so that seemed to simplify things, and I still had the same statistical conclusions in the end. Thank you for all of your help--this inspired me to look at my data differently and I feel more confident in moving forward with these results.

------------------------------
Courtney Cole
------------------------------

Original Message:
Sent: Mon February 14, 2022 08:45 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

My preference is to begin with simpler models and move to more complex only if needed. I don't know your hypothesis but as I see it here, simpler is better.

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

Original Message:
Sent: Mon February 14, 2022 08:04 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

I couldn't help but try running the data with the condensed categories for semester, and it appears to be working better:

It looks like there was just too much going on in my model from before...

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 07:31 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thanks again for responding!
I did not find any error when I ran these the first time. However, I did take your suggestion for Instructor and re-ran it (no more issues with instructor):

However, Semester still seems to be problematic. My next thought is to combine summer semesters together and combine spring and fall together (summer was condensed in comparison to the regular school year). I think it might just be a matter of too many "dummy variables" going on for this kind of analysis. Would you say that may be a fair assumption?

Thank you!

Courtney

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 06:23 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

I think I might leave Semester and recode instructor to 0-1 (Instructor=1 or not).

Did you have any warning messages when you ran these analyses?

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

Original Message:
Sent: Mon February 14, 2022 05:49 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Out of curiosity, I re-ran the analysis and removed semester from it completely. It hasn't fixed the problem, unfortunately. Could it really be a matter of unequal sample sizes?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 05:41 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hi Rick! Thank you very much for your response! I tried graphing counts of the DV by both instructor and semester separately. I can't really see what might be driving this issue (maybe something with unequal sample sizes from the perspective of instructor?). I know that instructors are technically nested in some of the semesters (Instructor 1 is the only one who is present in every semester, and the other three instructors only appeared in one semester each--so you can tell that Instructor 1 is the main person who teaches this course, but happens to share it with others from time to time)...so maybe I should just remove semester?

------------------------------
Courtney Cole

Original Message:
Sent: Mon February 14, 2022 02:52 PM
From: Rick Marcantonio
Subject: Multiple reference categories set to zero: How to interpret?

Hi, Courtney. The first thing I'd want you to look at are the counts and distribution of the DVs for those cells with the superscript "redundant" that are not the reference category for that factor (like Semester 1 and Instructor 3). The other thing I would be very curious about is that extreme value for beta(Instructor 2) in the second figure. Again, I'm no expert, but that's not passing the "smell test." ;)

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

Original Message:
Sent: Mon February 14, 2022 02:25 PM
From: Courtney Cole
Subject: Multiple reference categories set to zero: How to interpret?

Hello all!
I apologize for the many posts here lately, but I'm hoping this is relatively quick to answer.

I ran a generalized estimating equation and had not just one, but two reference categories:

I would simply like to know how to interpret this/state that it is okay to move on. These are control variables, but they were actually significant, so I am keeping them in the model. I saw a YouTube video (SPSS Generalized Estimating Equations) where the same thing happened there, but it was not explained.

I ran a multinomial logistic regression for another analysis and a similar thing happened:

Thank you!

Courtney

------------------------------
Courtney Cole
------------------------------

#SPSSStatistics