Analysing Likert scale outcome variable and how to combine questions into one

[Abraham Oluka]

Hello friends
I have a challenge analysing likert scale outcome variable. Allow me share full details of my research for more intuitive and adequate support

Main objective: "To determine the contribution of NGO capacity to Health Policy and System Research performance''

I have a total of 9 questions in my questionnaire focusing on my outcome above. Responses are recorded as 1-strongly agree, 2-agree, 3-Not sure, 4-Disagree, 5-Strongly disagree

Example of questions I have for my outcome variable

1. Our unit (including an individual staff) had a publication (s) of a research article (s) in a peer reviewed journal between September 2017 and August 2021
2. The unit (including individual staff) wrote an article (s) related to health policy and systems research in a national or regional newspaper from September 2017-August 2021.
3. I am happy with the research performance of my unit
4. Unit always uses research data to inform decisions in finance and programs/projects management
Note: (I have a total of nine questions for the outcome variable)

My independent variables (which I may not post here) are equally in likert scale responses 1-strongly agree, 2-agree, 3-Not sure, 4-Disagree, 5-Strongly disagree


What's the question then?
1. What's the most suitable statistical model to explain my outcome (though i was thinking of ordered logistic regression)
2. How do I combine the ten questions to explain my outcome rather than treating each question as stand-alone questions. OR is it statistically okay to combine questions into one to represent and outcome?

Thank you friends!

[Rhys Williams]

An ordered logistic regression sounds reasonable to me (treating your explanatory variables as factor variables (i.var notation)). However, I don't see why you wouldn't want to treat the explanatory variables as separate variables? That way you can see the effect of each explanation.
If you are looking for alternatives, then you could use something like principal components analysis, factor analysis or a simple Borda score (assigning the 1-5 numbers, add these up and divide by the total number of questions (10)).