Three developments in the statistical modelling of social survey data: as assessment of technical and substantive merits -PhD Thesis-
This thesis introduces and conducts three evaluations of statistical modelling techniques which are suggested to be useful additions for social science research, in this case with a particular focus upon the analysis of social stratification. The first issue concerns the prediction of income levels of the employed through Human Capital style models, whereby it is asserted that selection to the occupational position is a mechanism related to the income determination process in a non-ignorable way. The second issue concerns the possibility that when individuals are surveyed by sampling techniques which include household sharers, the non-independence between sharing respondents will bias the results of models which work at the individual level. The third issue concerns an argument that the relationship between ethnicity and social stratification may be well summarised by quantitative scores assigned to ethnic group categories, rather than accounts of ethnicity through dummy variable indicators or through models on separate ethnic group populations.
The analyses are conducted predominantly using two national survey resources, the British Household Panel Study (BHPS) and the Family and Working Lives Survey (FWLS). The discussion is thus oriented around dealing with accessible statistical methods at the level of survey analysis as communicated to social science research. The bulk of our evaluation involves comparing the results from reasonably simple statistical models which do and do not cater for the technique under discussion, over a set of permutations of relevant sociological concepts, variables and populations of study.
not held in Res Lib - bibliographic reference only