Table selection and log-linear models

Relationships between exposure to additional fluoride, social background and dental health in 7-year-old children

The relationship between social background, water fluoridation, use of fluoride tablets and dmf scores was studied in a birth cohort of 7-yr-old New Zealand children. This analysis showed that the level of exposure to additional fluoride was a complex variable influenced by at least three factors: the use of fluoride toothpaste; the child's length of residence in a fluoridated area; the length of time for which the child had been provided fluoride tablets. Log linear modelling of the data suggested the presence of a relatively complex relationship between exposure to additional fluoride, social background and dental health. First, both duration of residence in a fluoridated area and the provision of fluoride tablets made significant net contributions to the variability in the dmf score. However, exposure to additional fluoride from these sources was differentially influenced by social background: while there were relatively strong associations between social background and the use of fluoride tablets, there was no significant association between duration of residence in a fluoridated area and social background. Finally, independently of an association with exposure to additional fluoride, family social background made a significant net contribution to the variability in dmf scores.

Parental smoking and respiratory illness during early childhood: a six-year longitudinal study

The relationship between parental smoking habits and lower respiratory illness and symptoms during the first 6 years of life was studied in a birth cohort of New Zealand children. This showed that maternal (but not paternal) smoking was associated with significant increase in rates of lower respiratory infection and lower respiratory symptoms during the child's first 2 years. This association persisted when a range of perinatal, social, and familial factors were taken into account statistically. After two years there was no detectable association between parental smoking habits and lower respiratory infection. Further, there was no evidence to suggest that children whose parents smoked had increased risks of asthma or rates of asthmatic attacks during early childhood.

A categorical data analysis of contacts with the Family Health Clinic, Calabar, Nigeria

The relationships of population, environmental and accessibility variables to registration and attendance by mothers of children under 6 at the Family Health Clinic in Calabar, Nigeria are investigated. The technique used to analyze the data collected is categorical data analysis which proceeds in two stages, variable selection to reduce the variable set and fitting a log-linear model to the reduced set. Details of the statistical procedures used are provided to indicate how categorical data analysis can be used as a valuable tool of analysis in medical geographical studies that employ count or frequency data. It was found that younger mothers and Ibibio women registered more often at the clinic than did their counterparts. However, if the relatively sparse data on fathers is accepted, the association between age and registration is found to be spurious and a model can be substituted which shows younger fathers and fathers who spoke a non-Efik/Ibibio language to be associated with higher clinic registration of mothers. It was further found that for registered mothers the probability of a clinic visit was decreased by mother's age, increased by distance given no travel cost, unaffected by distance given some travel cost, increased by travel cost given a short distance to the clinic and decreased by travel cost given a longer distance from the clinic. These results are discussed in relation to population characteristics such as socio-economic status, clinic procedures such as health worker activities, transportation availability in Calabar, the spatial ecology of the city and local environmental conditions.

Multivariate logit analysis of concordance ratios for quantitative traits in twin studies

The application of multiway contingency table analysis to the multivariate analysis of concordance ratios in twin studies is developed. The approach is illustrated by data on smoking and alcohol use in Finland and Sweden. This approach can enable the assessment of the effect of other variables on the concordance ratio and thus allow estimates of genetic effects on the trait under study. Hypotheses on relationships between genetic effects and other variables can be tested. After hypothesis testing, model fitting of the best hypothesis can be carried out.