Correspondence analysis and logistic medelling: Complementary use in the analysis of a health survey among nurses

Abstract

Logistic or other modelling approaches are often appropriate for studying epidemiological data. However, health surveys may be more complex. With numerous variables, there is a need for exploratory analysis. Parsimonious description of the data is also a useful complement to modelling. Correspondence analysis may be useful in such exploratory phases. An example of the complementary use of the two approaches is presented in the context of a health survey among nurses, which focused on the relationship between shift and night work, and declared health problems.