1. Invited talk by David Bellhouse, Department of Statistics, University of Western Ontario: "Developments in the Analysis of Complex Surveys within the Canadian Statistical Community."

1. Martin Cooke, Department of Sociology, UWO. "Social Assistance and Female Lone Parents: An Analysis of Durations and Transitions."
   
   
2. Byron Lee, Department of Economics, University of Waterloo. "Educational Efficiency and Returns to Education."
   
   
3. Amanda Franceschini, Department of Family Relations, University of Guelph. "A Longitudinal Study of Family Relationships as Mediators of Children's Schooling Outcomes."
   
   
4. Luc Boyer, Department of Sociology, UW. "Mapping Non-ignorable Nonresponse Behaviour."
   

12 - 1:15 Lunch

Chair of Afternoon Session: Keith Warriner, SWORDC Co-Director

1:30 - Session on works in progress

1. Zenaida Ravenera and Rajulton Fernando, Population Studies Centre, UWO. "Family Transformation and Social Cohesion Projects."
   
   
2. Pat Newcombe-Welch, SWORDC, and Tim Seifert, Faculty of Education, Memorial University. "Big Country, Small Sample - What can we safely conclude?"
   
   
3. John Goyder, UW Sociology, M.E. Thompson, Statistics and Actuarial Science, and Shane Dixon, Sociology. "Scaling the Major Group of the National Occupational Classification."

- - - - - - -- - - - - - - - - Coffee Break - - - - - - - - - - - - - - - - - - -

3:00 - Roundtable Session on the RDC Application Process.

(A outline of how the process works, and exchange of experiences, tips and questions about getting approval to conduct an RDC project).

ABSTRACTS:

Note: Not all participants submitted abstracts.

Morning Session, Presententation Number 2
Byron Lee,
Department of Economics
University of Waterloo

Title: Educational Efficiency and Returns to Education.

Abstract:

This paper analyzes the relationship between educational efficiency represented by the time taken of the individual above or below the normal time taken to complete the degree and the returns to education. It includes controls for the experience, years of schooling, education credential level, field of study, occupation other demographic variables and geographical variables. The model clearly shows that there are earning gains with an increase in educational efficiency. Previous models have failed to to measure the unobserved ability of the individual. The resulting coefficients on schooling and credentials may be more accurate as it takes into account this potential bias. Educational efficiency is assumed to be a signal (not taught through a university education). As a result, since educational efficiency is positively correlated with returns to education, this shows that it is (at least partially) through this signal that earnings are increased.

Morning Session, Presententation Number 3
Amanda Franceschini
Department of Family Relations
University of Guelph

Title: A Longitudinal Study of Family Relationships as Mediators of Children's Schooling Outcomes.

Abstract:

It is generally accepted that both the family and the school play major roles in fostering the optimal learning and social development of children and youth, with implications for later well-being. However, in a climate of seemingly increasing demands on time, decreasing resources, and increasing pressure to produce measurable results, parents, educators, and other professionals may find themselves questioning how to effectively nurture children's success in school.

The objective of this study is to examine within-family processes as mediators of children's schooling outcomes. Specifically, positive and hostile- ineffective parent-child interactions, parental support related to school, and children's general self-concept, academic focus and pro-social characteristics are predicted to affect children's academic achievement and social adjustment.

The Family-School Relationships Model (Ryan & Adams, 1995) will be used as a theoretical guide to investigate predictor and outcome relationships and individual differences in the effects of these mediating variables. With cycles of teacher, parent, and child self-report responses available two years apart, the National Longitudinal Survey of Children and Youth (N.L.S.C.Y.) will be used as a source of Canadian data. Namely, cohorts from Cycles 1 and 3 will be examined. Structural equation modelling will be applied as a statistical tool.

Morning Session, Presententation Number 4
Luc Boyer
Department of Sociology
University of Waterloo

Title: Mapping Non-ignorable Nonresponse Behaviour.

Abstract:

Nonignorable nonresponse occurs if any value of a survey variable of interest correlates with nonresponse behaviour, or if the value of a survey variable of interest and the nonresponse behaviour correlate to a common variable (internal or external to the survey). Such conditions will generate a nonresponse error, whereby the respondents' distribution for the survey variable will differ from that of the population. Perhaps due to the concern for inferential validity in a declining response rate environment, researchers are now paying closer attention to the relationship between nonresponse and nonresponse error, than to nonresponse alone. Empirical investigations, still very limited in number, are nonetheless predominantly conducted within the methodologists' or the experimentalists' framework. While nonresponse bias analyses or survey design impact analyses are useful descriptions of nonresponse error, such investigations are neither able to explain its occurrence, nor establish its relationship to nonresponse behaviour. As a consequence, they offer little insight into potential nonignorable nonresponse mechanisms. Using the Waterloo Graduate Student Exit Survey, in conjunction with Statistics Canada's Survey of Graduates dataset, this exploratory case study will endeavour to generate hypotheses that account for the relationship between students' satisfaction levels, and their response behaviour to the exit survey. By using Michalos' discrepancy theory to identify potential specification or explanatory variables, the goal is to uncover nonignorable nonresponse behaviour mechanisms.

Afternoon Session, Presententation Number 2
Pat Newcombe-Welch,
Statistics Canada,
South Western Ontario Research Data Centre,
University of Waterloo,

Tim Seifert,
Faculty of Education,
Memorial University

Title: Big Country, Small Sample - What can we safely conclude?

Abstract:

Statistics Canada provides survey weights along with its data files and advises that the weights be used in order to produce unbiased estimates. The survey weights are constructed so that known subtotals, such as provincial populations and gender totals are preserved. One aspect of this method is that the sample respondents from a small province such as Prince Edward Island will on average have much smaller weights than the sample respondents from a large province such as Ontario. If a national statistic is based on a subsample which contains relatively few respondents from each province, although the weighted estimate will be theoretically unbiased, the accompanying variance may be expected to be large due to the large variability amongst the weights. Researchers who experience this phenomenon may feel justified in using an unweighted analysis. An example of this type from the National Longitudinal Survey of Children and Youth (NLSCY) is presented, in which it is demonstrated that it is possible to conduct a weighted analysis and minimize the effect of variance inflation by using provincial groupings as a variable in the model.

Afternoon Session, Presententation Number 3
John Goyder, M.E. Thompson and Shane Dixon
University of Waterloo
(Departments of Sociology, Statistics and Actuarial Science, and Sociology respectively)

Title: Scaling the Major Group of the National Occupational Classification.

Abstract:

Human Resources and Development Canada developed in the early 1990s a classification of occupations having a more sociologically sensitive major groups structure than older codes such as the CCDO of the 1970s. This National Occupational Classification was updated in 2001, but the system is not used as widely as might be expected. Researchers seem to want major groups that are ordered into a scale, such as the Pineo, Porter, McRoberts scale of 1977. In 2000, a study of occupational prestige was collected in Kitchener-Waterloo, Ontario, with personal interviews with 351 respondents. The survey included ratings of the 26 major groups of the NOC. The paper describes the construction and validation of the scale. The resulting scale for NOC major groups will be especially useful for researchers wishing to scale the prestige of occupations on public use versions of Statistics Canada data sets. Such files normally report only the first two digits (i.e., the major group) of occupational code.