Abstract This work presents the results from an application of multiple procedures of data analysis to a body of data consisting of the life history information gathered from next of kin listed on death certificates of 500 white male deaths in Boston during 1965. Seventy variables were analyzed by zero-order correlation, multiple regression, linear discriminant analysis, factor analysis, nonlinear discriminant analysis, and nonlinear clustering. The results show a substantial shift in significance of independent variables as one proceeds from univariate to multivariate and from linear to nonlinear analysis. Various perturbations in such studies are outlined, such as the effects of secularity and bimodal distributions. The substantive findings suggest that physical predictors are more important than social predictors to account for advanced longevity while the contrary holds for moderate longevity. Although social variables are more vulnerable to secularity than physical variables (which whittle down their presence in the variable set and consequently their ability to emerge as important relative to physical variables), there was some evidence of their being more important than physical variables.
[1]
B. Bell,et al.
The Veterans Administration longitudinal study of healthy aging.
,
1966,
The Gerontologist.
[2]
M. B. Wilk,et al.
Data analysis and statistics: an expository overview
,
1966,
AFIPS '66 (Fall).
[3]
C. L. Rose.
SOCIAL FACTORS IN LONGEVITY.
,
1964,
The Gerontologist.
[4]
O. J. Dunn,et al.
Elimination of variates in linear discrimination problems.
,
1966,
Biometrics.
[5]
G. N. Lance,et al.
A General Theory of Classificatory Sorting Strategies: 1. Hierarchical Systems
,
1967,
Comput. J..
[6]
Resolution of the heywood case in the minres solution
,
1966
.
[7]
H. Harman,et al.
Factor analysis by minimizing residuals (minres)
,
1966,
Psychometrika.
[8]
K Enslein,et al.
A general-purpose perceptron simulator.
,
1967,
Computers and biomedical research, an international journal.