Open AccessMethod: The New Zealand Socio‐economic Index of Occupational Status: methodological revision and imputation for missing data
Author(s) -
Davis Peter,
Jenkin Gabrielle,
Coope Pat,
Blakely Tony,
Sporle Andrew,
Kiro Cindy
Publication year - 2004
Publication title -
australian and new zealand journal of public health
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.946
H-Index - 76
eISSN - 1753-6405
pISSN - 1326-0200
DOI - 10.1111/j.1467-842x.2004.tb00922.x
Subject(s) - imputation (statistics) , logistic regression , discriminant function analysis , workforce , missing data , statistics , population , census , econometrics , geography , environmental health , medicine , mathematics , economics , economic growth
Objectives : To revise and update the New Zealand Socio‐economic Index (NZSEI) in the light of methodological issues in its construction, and to develop an imputation method for use where occupational information is not available. Methods : Data were drawn from the following New Zealand national surveys: 1996 Population Census; 1996/97 and 1997/98 Household Economic Surveys; 1996/97 Household Health Survey. Three sets of statistical analyses were applied: alternating least squares to generate socioeconomic scores; cluster and discriminant function analyses to identify cut‐points; and regression and logistic regression to develop and test imputation methods. Results : Socio‐economic scores for the full‐time workforce in 1996 showed a different distribution, but much the same occupational ordering, as in 1991. The introduction of part‐time workers and income adjustment multipliers for self‐employed workers significantly affected scores for management and agricultural titles. The application of cluster and discriminant function analyses generated six groupings that were relatively distinct occupationally. An imputation method based on an averaging of scores within age/qualification categories was found to achieve acceptable results. Conclusions : Methodological improvements in the construction of the NZSEI have enhanced its empirical robustness, while a simple imputation technique has widened the potential application of the scale.