The overall aim of this project was to determine whether or not there is a relationship between intellectual disability and poverty and health variables in Brazil and the UK. Through determining this relationship, the aim was to then try to profile a child who is at greater risk of intellectual disability in order to result in a quicker diagnosis and earlier access to the support and resources available for children with intellectual disabilities.

In order to investigate this relationship in Brazil, The Pesquisa Nacional de Sa'ude (PNS) was used. For the UK, data from the Millennium Cohort Study (MCS) was used.

In order to account for the complex survey design, both model-based and designbased approaches to analysis were investigated. Simulations were run to compare the various methods and recommendations about which methods to use in various scenarios were made.

Due to the large number of variables available in the two data sets, methods of
variable selection were examined. Both stepwise selection based on Akaike Information Criterion (AIC) and the lasso were compared through simulations. It was found that although these methods resulted in different models being selected, the inference made based on the selected models did not vary much between the two methods.

To conduct the analysis of the PNS and the MCS a design-based approach was taken. Stepwise selection using AIC was used for variable selection and sampling weights were used when calculating the coefficient estimates and standard errors.

After the analysis of the PNS data, a potential profile of a child likely to have an
intellectual disability in Brazil was found to be: a child who is unable to read and write with poor general health and multiple visits to doctor within a 12 month period. In the UK, it was found to be: a child in a family who requires extra support in the form of benefits along with a poorer general health which limits daily activities.