Human development classification from environmental indicators using LDA

The Sustainable Development Goals (SDG) can be monitored through statistical evaluation of indicators of interest. These indicators might measure social and environmental issues. But there is a lack of studies that assess the... [ view full abstract ]

The Sustainable Development Goals (SDG) can be monitored through statistical evaluation of indicators of interest. These indicators might measure social and environmental issues. But there is a lack of studies that assess the association between human development and environmental performance through the use of modern statistical learning techniques. The aim of this study was to estimate the association and a predictive model of the human development index from environmental performance indicators by the means of Linear Discriminant Analysis (LDA) and cross-validation (CV). The dataset was built merging the 2015 United Nations Human Development Index (HDI) and the 2016 Yale’s Environment Performance Index (EPI) datasets by country. It consisted of 76 observations of the following selected environmental indicators: population lacking access to sanitation (unitless), tree cover loss (%), health risk from household air quality (unitless), health risk from PM2.5 exposure (unitless), wastewater treatment level (%), exposure to unsafe water quality (%) and population lacking access to drinking water (unitless). A LDA was conducted considering the HDI classification as a response variable and the selected environmental indicators as explanatory variables. The Leave-One-Out-Cross-Validation (LOOCV) technique was used to evaluate the predictive ability of the model. All analyses were conducted using the R Software v.3.3.1. The first two estimated linear discriminant functions showed a cumulated explained variance equal to 97%. The built LDA model was able to correctly predict 84% of the sample observations. Considering the LOOCV technique was used, this results suggest there is a strong association between the well-being of a population and its environmental performance. This means that changes in the studied environmental indicators are linked with changes in a population’s well-being. It also suggests the model can predict with high accuracy the well-being of a nation from environmental variables. This tool can help scientists, politicians and managers to monitor and improve the performance of sustainability indicators.