CHAPTER 3 DATA COLLECTION AND ANALYSIS 1.2 Other Sources The variety of thematic areas covered by the Global Study required gathering country-level data on the number of children deprived of liberty from numerous sources. Although the priority has been always given to the data submitted under the Global Study questionnaire and extracted from the official registries (e.g. police records), the existing data gap was partially filled with the information reported by international organisations, most notably the UN agencies. These data sources were especially important for estimates in the areas of national security and armed conflict (States were unable to provide data). If there were still many values missing after the inclusion of these sources, then the next step involved extraction of relevant data from peer-reviewed literature. This has been done for the chapters on migration-related detention and children in institutions. Priority has been given to the data provided by public authorities and the conflicting outlying values have been excluded. In case of discrepancies between equally credible and timely information, the most conservative data were chosen. Thus, the figures presented in the Global Study are treated as minimum estimates. Application of regression-based methods required the collection of additional data (socalled predictors3) that allowed estimating figures for the countries for which recent data were not available or reliable. For this purpose, two kinds of predictors have been collected: • General variables (common across all thematic areas): geographical region (according to UNICEF regional classification), population (total population as well as population 0-14 and 0-19 extracted from World Bank/UN DESA), GDP per capita (extracted from the World Bank) and Gini index (inequality index; extracted from World Bank); • area-specific variables (for the purpose of building a model in a particular thematic area) such as the prison population rate and minimum age of criminal responsibility (in the field of the administration of justice), the international migrant stock and the Human Development Index (in the field of migration), fertility rate and female prison population (in the field of children living in prisons with their primary caregivers). Selection of predictors has been guided by the relevance (aimed at controlling for the target variables), availability of data (priority was given to data reported by public authorities and international organisations), completeness (predictors with high rate of missing values were excluded) and timeliness (preference was given to the most up-to-date information). 3 ‘Predictors’ are variables regarded as important factors in explaining a certain phenomenon, e.g. for assessing the number of children deprived of liberty in the administration of justice, important factors are inter alia the population of children in a particular country, minimum age of criminal responsibility or prison population rate (per 100,000 citizens). 38