Resilience and Deprivation in Welsh Children

The study primarily aimed to investigate the association between the resilience profile PLP derived by the study and the local area deprivation measured by WIMD among the children living in high household-level deprivation (FSM children). The current study, however, also investigated a similar association among the non-FSM children group, hence FSM-stratified analysis was performed. A supplementary analysis has discussed the interaction between FSM eligibility and WIMD. This study examined if a child's potential to leave poverty can be moderated by improvements in their in local built environment. This is measured by examining the association of the domains of WIMD (e.g. income, community safety, health, access to services, physical environment, housing) on a child's outcome in order to develop insight into the factors that best influence the child's trajectory. Logistic regression models were used to determine the association between local area deprivation measured by WIMD and achieving PLP amongst the children in Wales. The logistic regressor was augmented with stepwise bidirectional (forward and backward) search for optimal model selection (Burnham & Anderson, 2003 ). This method determines the best model with the minimum Akaike Information Criterion (AIC) and least significant features are excluded at each iteration step. The study has confirmed that there is no major concern around the high degree of correlation between predictor variables in the regression models by multicollinearity test (see Supplementary material collinearity test). Along with the explanatory variables, the stepwise logistic regression models have been adjusted for other covariates – such as exam year, gender, urban/rural classification of the living area, number of adults in the household, number of children in the household, living with someone who had an alcohol problem, living with someone who had depression, living with someone who had serious mental illness, child's special education need requirement – as these factors are also associated with the outcome variable. The odds ratio calculated with this adjustment has been reported throughout this work. The statistical significance of the explanatory variables and covariates have been interpreted by the p value less than 0.05. The data preparation including extraction, cleaning and linkage was performed in Structured Query Language (SQL) on an IBM DB2 platform and analyses were performed in the R statistical language version 3.3.2 (R Core Team, 2018 ).