Please what is the possibility or how do one go about using indicators that are ratio or nominal in nature or a combination of both to load onto a latent variable. In this case the objective is not dimension reduction but just to be able to arrive at a single figure that will represent the predictor(latent variable) which can then be used to regress the outcome variable. E.g opportunity(predictor) is a latent variable(not measurable) but instead of using a questionnaire, for questions relating to opportunity, I have a combination of indicators of ratio and nominal observed variables from which I would like to derive a single figure of opportunity to use as predictor in my regression equation.
Dear Murumba
I don't follow everything you have said in your question but, in general terms, the problem data is the nominal variables. These usually need to be broken down into multiple binary variables, known as dummy variables. The precise way you combine everything together into a single latent variable will depend on your technique.
Hi Peter, thanks. What am saying is that where we have predictor variables as latent variables which are not measurable. What is done in most cases is to collect data via questionnaire and do factor analysis to arrive at the loading of the indicator variables on the latent variable. With that you obtain measurable value of the latent variable which can then be regressed with the outcome variable. My question is that where you are not using questionnaire but have observed variables comprising ratio and nominal scale, can factor analysis be used to arrive at factor loading unto the latent variable or another technique of analysis is appropriate or is not possible at all.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression