Both of my independent variable and dependent variable use 5 likert scale. Each variable contains 8-10 items/questions. What is a suitable analysis for this data set?
Noor Syafinas Muda maybe this article may clarify this issue
http://hosted.jalt.org/test/bro_34.htm#:~:text=Interval%20scales%20show%20the%20order,the%20points%20on%20the%20scale.&text=Scales%20based%20on%20Likert%20items,scale%20make%20sense%20as%20ratios.
You can perform a predictive analysis from VI to VD, you can also see if this depends on or varies according to a categorical variable (sex, age) according to your theory.
Cristian Ramos-Vera Thank you for you response. I'm thinking to use multiple linear regression but I'm not sure if it is acceptable to compute mean of each variable and treat it as continuous data.
If this type of analysis is okay, it is used as a continuous variable.
My suggestion would be ordinal logistic regression. See the attached notes. Best, D. Booth
David Eugene Booth Thanks a lot for your suggestion and the notes. In order to do ordinal logistic regression, then I have to calculate a composite score for median instead of mean, right?
David Eugene Booth Thanks for you response. I used likert scale which composed a series of 8 likert-type items for both my IV and DV. So, in order to do the regression analysis, I have to combine the likert-type items for each variable into a single composite score/variable. If I do linear regression, then I have to calculate a mean as a single composite score. But in order to do ordinal logistic regression, then I have to calculate for its median. Is my understanding is correct?
Noor Syafinas Muda Why do you think that you need to do something different for ordinal logistic regression?. The only thing different should be that this is the appropriate regression method for ordinal DVs. This should have nothing to do with scale construction as far as I can see only with the properties of your final DV. See this link:
https://www.google.com/search?client=firefox-b-1-d&q=scale+construction+from+likert+questions+ResearchGate
If you are still unsure, please post a question about that.
Best wishes, David Booth
David Eugene Booth I've read about the argument of treating ordinal data as continuous and I agree with you. Since my DV is ordinal, I should use ordinal logistic regression. Thank you. Much appreciate!
Noor Syafinas Muda maybe this article may clarify this issue
http://hosted.jalt.org/test/bro_34.htm#:~:text=Interval%20scales%20show%20the%20order,the%20points%20on%20the%20scale.&text=Scales%20based%20on%20Likert%20items,scale%20make%20sense%20as%20ratios.
- Badges
- Science topic