Hi Helen
As I understand your question then you might want to use binary logistic regression, with two dummy coded variables (medium vs low) and (high vs low) predicting your outcome variable of yes/no in block 2....following the control variables already entered in block 1. Although granted you would have to do a separate logistic regression for each question (could be problematic if you've got loads of questions). I don't often deal with data that is categorical but assuming you have quite a few questions, is this design not most suitable for a pearson chi-square test?
Regards
Tom
Tom Bailey BSc MSc
PhD Researcher at The University of Nottingham
Blog - http://tomjamesbailey.wordpress.com
LinkedIn - uk.linkedin.com/in/tomjbailey7
Twitter - https://twitter.com/TomJBailey7
________________________________________
From: Research of postgraduate psychologists. [[log in to unmask]] On Behalf Of Jeremy Miles [[log in to unmask]]
Sent: 16 October 2014 16:43
To: [log in to unmask]
Subject: Re: Multinomial Logistic Regression
When you add more predictors, you control for those other predictors. If that's what you want to do, that's what you should do. (That is interfering with the model, because it changes the model - but that might be the model you want).
You can still look at interactions if you want to.
I'm not sure what it's saying about model fit, 'cos I don't use SPSS . But I would have thought you wanted significant model fit.
Also, why aren't you doing ordinal logistic regression? Your outcome looks ordinal, rather than categorical to me.
There's a nice book on all this by J Scott Long called "Regression analysis for categorical and limited dependent variables."
You say "What I'm interested in mainly is does group membership affect their yes/no answer". So is yes/no answer your outcome? I assumed it was group membership. If it's yes/no, you only have two options, and regular binary logistic regression will work.
Jeremy
On 15 October 2014 09:27, Helen Mann <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hi,
I have three groups low/medium/high and their yes/no answer to questions.
If I do a multinomial logistic regression for each question and lump in other things such as "age", "gender" etc. does that interfere with my model (like it would if you added things in a normal linear regression?). As I have loads of IV's I thought I would lump them all into the one analysis to save time - thinking that as it's all to do with odds ratios that there would be no looking for interactions between the IV's on the DV (e.g. if group and gender was there it would look at group on answer and gender on answer but not affect of group and gender on answer)...I have found this from Strathclyde University (https://www.strath.ac.uk/aer/materials/5furtherquantitativeresearchdesignandanalysis/unit6/multinomiallogisticregression/) and have been trying to follow it but I think what is confusing me is this idea of model fit...what does it mean by model fit??? Why do I want the model likelihood ratio test to be non significant.
What I'm interested in mainly is does group membership affect their yes/no answer - does it complicate things to add in other variables? If I had continuous data I would run an ANOVA but alas its categorical...
Help!
Helen
This message and any attachment are intended solely for the addressee and may contain confidential information. If you have received this message in error, please send it back to me, and immediately delete it. Please do not use, copy or disclose the information contained in this message or in any attachment. Any views or opinions expressed by the author of this email do not necessarily reflect the views of the University of Nottingham.
This message has been checked for viruses but the contents of an attachment
may still contain software viruses which could damage your computer system, you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation.
|