I have recently started to improve my methodological skills and programming in R. For a study term paper I now want to run an ordered logit/probit model. The data I use are taken from the European Social Survey, which I have merged over all available years, and built subsets containing the variables of interest, that are:
Dependent variable -> “happiness” (ranging from 0 to 10, with 0 = “Extremely unhappy” and 10 = “Extremely happy”) Independent variables -> “Income” (ranging from 1 to 10), “health” (ranging from 1 to 5), “religion” (ranging from 0 to 10) and “personal relationships” (ranging from 1 to 7)
Of main interest is the relationship between “happiness” and “income”, however, I want to control for further factors.
My research so far suggests that an ordered logit/probit regression model is the most appropriate here. I have read some literature regarding this method, and am not sure about some points.
- As far as I understand, a latent variable that is not observable influences the answering process and thereby the model and its results. Do I need to program this latent variable manually, or is it “automatically” done by R (I have used the MASS package so far, but oglmx was further suggested to me, to control for heterogeneiety)?
- I am not sure about the “threshold parameters”. Again, do I need to program them as well, or is it done automatically? And what exactly are they good for?
- Furthermore, I am interested in, how exactly to interpret the outcome. Is it possible to interpret the coefficients at all, or just the fore sign?
Additionally to my questions, I am very open to all sorts of tips regarding this model (or more appropriate ones?), due to the fact that I am new to this and want to learn.