Child hunger in Canada: results of the 1994 National Longitudinal Survey of Children and Youth
Murray Finkelstein
Email: murray.finkelstein@utoronto.ca
Affiliation: Mt Sinai Hospital, Toronto
Posted on: 10/20/00
Lynn McIntyre and colleagues have analyzed data from the National Longitudinal Survey of Children and Youth and reported on child hunger in Canada. Unfortunately, their methodology does not permit them to extend their conclusions beyond the sample they analyzed to the population of Canada. The sample in the National Longitudinal Survey of Children and Youth is not a random sample of Canadian children. It becomes representative of the
population of Canada only when analysts take account of the sampling weights provided by Statistics Canada. McIntyre and colleagues state that they did not use the sampling weights because they were not generating population estimates. While their results might provide some information about child hunger in Canada, there is no guarantee that they are representative of the
country as a whole. The title of the paper and editorial comment should thus have been "Child Hunger in a Sample of Canadian Families".
When working with survey data, analysts will usually want to generate inferences on the Canadian population, rather than limit themselves to the subjects of the survey sample. This can be accomplished by making use of the sampling weights provided by Statistics Canada in all analyses. It is more complicated to take account of the Complex Survey Design and to correctly
compute Confidence Intervals. This is generally now done by using boostrap methodology.
Murray M. Finkelstein
Department of Family and Community Medicine
University of Toronto
Copyright 2000 Canadian Medical Association or its licensors
Canadian
Medical Association Journal