Hi
I am struggling with deciding the correct analysis method for demographic variables on a number of DVs. I have seen mulitnominal logistic regression but not sure if that would need a categorical DV?
I have completed my main analysis but I want to check the effect that the demographic data has on the DVs.
The demographics are mostly categorical e.g. age range categories, gender has 4 options (male, female, non-binary, prefer not to say), Type of school has several options etc.
The DVs are all scales using likert scales either 1 - 5 or 1 - 7 and I am using the total mean for each one so ordinal level.
Hope that makes sense!
Thank you for your help,
I was advised to run chi-square and then compare the means but given the ordinal level data of DV I am not sure this is the best option.
Your DVs are ordinal so you need multivariate ordinal logistic regression. Believe it or not if you google this term and your favorite stat package you will find descriptions of the methods to do it. Videos too. Best wishes, David Booth PS if you have only one DV I wasn't sure you should use ordinal logistic regression. See the attachment for an introduction.
For your DVs, there is an important difference between individual, Likert-scored items and scales that are composed of multiple such items. If you have multi-item scales, then it is possible that these can be converted into interval-level variables.
David L Morgan Thank you for this. I was having this discussion today funnily enough. As I understand it, the scales are validated scales so I think can be treated at interval level on the basis that they have been tested on a range of normally distributed large samples. I was advised instead to use MANOVAs which I did and found just one participant characteristic (whether a school has a positive education initiative) to effect the main DV (Team Learning). But in my sample of 108 teachers (G*Power was min. 85), most of the schools had a pos. ed programme. I added it as a first step in the main hierarchical multiple regression and it gave an R sq of .089. which is teeny but significant. I'm not sure I took the right approach here!
Yes, it sounds like you are taking the right steps.
Just to be careful, I would look at Coefficient Alpha for each of your Dvs.
David L Morgan Thanks for your reply. I previously checked alphas for the scales (all above .80) and I checked VIFs etc for multicollinearity. Thanks for your help today
- Badges
- Science topic
- Similar topics
- Statistics
- Survey