Living Conditions, Interviewer Effects and Perceived Well-Being of the Elderly. A Multiple Correspondence Analysis Approach.

This paper reports on a study on quality of life of elderly people carried
out in the city of Girona (Spain) in 1999. The study of the quality of life of
the elderly must be based on both objective and subjective indicators along
a set of relevant sub-dimensions. Most of the relevant factual and subjective
items in quality of life questionnaires are qualitative and call for a multiple
correspondence type of analysis. Besides, most of the questions are to some
extent sensitive and therefore prone to high non-response and interviewer
effects.
In this paper, drawing on the work of Escofier (1981) and Zárraga and
Goitisolo (1999), we apply a variant of multiple correspondence analysis
that can be implemented with ordinary principal component analysis
software and that prevents non-response categories from having too high a
contribution on the first dimensions. Subjective well-being questions play
the role of active variables and objective well-being questions that of
illustrative variables. Next, analysis of variance models are fitted to the axis
scores with the interviewer and demographic variables used as predictors.
Interviewer effect estimates are used to partial interviewer effects out of the
axis scores.
The results show a two-dimensional solution to be appropriate. The
upper right quadrant corresponds to high quality of life and the lower left
quadrant to low quality of life. The solution is related in the expected way
to many of the objective illustrative variables such as neighbourhood, prior
occupation, income source, disablement, education, level of physical
activity and housing condition.
The analysis was replicated without accounting for non-response and
interviewer effects and the interpretation of the axes became much less
clear.