\ Global Youth Development Index and Report 2016 Figure A2.1  The proportion of data available and imputation methods of the YDI 40% PROPORTION Figure A1.2 shows the proportion of data available and gives an imputation summary of the YDI between 1990 and 2015. Of the original data used, 96 per cent are later than 2010 and no data before 1990 are included. The construction of the YDI makes the best use of all data available to estimate the level of youth development. However, caution must be exercised when analysing YDI scores for a country over time. The large drop in original data in the years between 2010 and 2015 shows that sometimes countries can have just a single data point in the time period for one or more indicators. 30% 20% 10% 0% No Imputation Interpolated Extrapolated IMPUTATION METHOD Mean Imputation Source: Commonwealth Secretariat In using hot and cold deck imputation methods, the YDI uses the best possible data without an overly complex methodology. Countries with more than 50 per cent of data missing are not included in the YDI. On average 87 per cent of the YDI score of remaining countries is based on existing country data imputed by either time series or cold deck techniques. Some 88 per cent of the 2016 YDI is either original data or interpolated or extrapolated from original data. The remaining 12 per cent is based on hot deck imputation techniques. Figure A2.1 summarises data availability and imputation methods of the YDI. One of the findings in this report is that youth development has improved since 2010. In full, this should be interpreted as ‘given the availability of new information for each year, the evidence suggests that youth development has improved since 2010’. The paucity of data prohibits anything stronger being claimed as a result and highlights that collecting more and better age-disaggregated data needs to be made a priority to further understand country and global trends. A2.4  The banding process In order to allow aggregation of noncommensurable indicators, a process of normalisation or banding was undertaken. Under this process each indicator is scaled between a score between 0 and 1, relative to a global dataset. To do this, appropriate minimum and maximum Figure A2.2  The proportion of data available and timeliness of data to YDI year, 1990–2015 100% 75% Proportion 142 50% 25% 0% 1990 Source: Commonwealth Secretariat 1995 2000 2005 Timelines of data to YDI year 2010 2015