Hi everyone,

I'm new to SPSS so still learning. I am trying to conduct a regression analysis. I have got a scale dependent variable (Student grades) and a few independent variables, one of which is ordinal (10 Likert scale questions) . What regression should I use? Ordinal logistic regression?

Thanks!

------------------------------
Aaron Price
------------------------------

#SPSSStatistics

Hi. It may be alright to treat that ordinal variable as continuous, but my advice is that you read this before you do anything.

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

I would disagree with that reference where the author reasons that just because we don't know that scale variable effects are really linear, it's okay to treat ordinal variables as continuous (oversimplifying a bit).

Ordinal regression, PLUM in Statistics, is for ordinal dependent variables, not ordinal predictors.  But it may be appropriate to treat the ordinal variables as factors and choose appropriate contrasts with them using ordinary regression.

If the variables of interest are not the ordinal ones, running the regression with ordinals as factors and ordinals as linear covariates would allow you to see whether the ordinality assumption matters.  If it doesn't, you are home free.

------------------------------
Jon Peck
------------------------------

Original Message:
Sent: Tue April 05, 2022 09:26 AM
From: Rick Marcantonio
Subject: Regression

Hi. It may be alright to treat that ordinal variable as continuous, but my advice is that you read this before you do anything.

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

Ok thanks for the help Jon, I'll try it out.

------------------------------
Aaron Price
------------------------------

Original Message:
Sent: Tue April 05, 2022 11:25 AM
From: Jon Peck
Subject: Regression

I would disagree with that reference where the author reasons that just because we don't know that scale variable effects are really linear, it's okay to treat ordinal variables as continuous (oversimplifying a bit).

Ordinal regression, PLUM in Statistics, is for ordinal dependent variables, not ordinal predictors.  But it may be appropriate to treat the ordinal variables as factors and choose appropriate contrasts with them using ordinary regression.

If the variables of interest are not the ordinal ones, running the regression with ordinals as factors and ordinals as linear covariates would allow you to see whether the ordinality assumption matters.  If it doesn't, you are home free.

------------------------------
Jon Peck
------------------------------


Original Message:
Sent: Tue April 05, 2022 09:26 AM
From: Rick Marcantonio
Subject: Regression

Hi. It may be alright to treat that ordinal variable as continuous, but my advice is that you read this before you do anything.

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

Great! Thanks, Rick, I'll give it a read!

------------------------------
Aaron Price
------------------------------

Original Message:
Sent: Tue April 05, 2022 09:26 AM
From: Rick Marcantonio
Subject: Regression

Hi. It may be alright to treat that ordinal variable as continuous, but my advice is that you read this before you do anything.

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

Look also this https://stats.stackexchange.com/q/195246/3277

------------------------------
Kirill Orlov
------------------------------

Thanks Kirill!

------------------------------
Aaron Price
------------------------------

Original Message:
Sent: Tue April 05, 2022 06:07 PM
From: Kirill Orlov
Subject: Regression

Look also this https://stats.stackexchange.com/q/195246/3277

------------------------------
Kirill Orlov
------------------------------