Abstract
In this paper, we study the connections of categorical levels of Human Development Index (HDI), GDP per capita, World Happiness Index, the Gini indexes of Income and Wealth inequalities together with poverty rate for 98 world countries. By clustering analysis, we identify four groups of countries with similar features. K-means clustering algorithm is applied to obtain four clusters of sizes 21-26 countries by explaining 68.3% of the total variation in data. The analysis reveals significant differences between the clusters, while also factors with largest differences within the clusters. Secondly, multinomial logistic regression (MLR) is applied in predicting the HDI categories of the full sample of 98 world countries for year 2018. The MLR model can capture also nonlinear relationship. The logistic regression model achieved 91.8% overall accuracy. The results of our research together from earlier literature is followed by suggestions for the future research.
Highlights
We concluded that the multinomial logistic regression classifier was somewhat more efficient, with an accuracy 92.9% compared with the discriminant analysis accuracy of 83.7%
881 In Table 7, likelihood ratio tests show that the intercept, GDP per capita, Poverty rate and Wealth Gini have p-values less than 0,10; they have a significant contribution to the full effect. Of these statistically significant components, Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) for the reduced model shows the lowest values for Wealth Gini and Poverty rates together with Intercept
The clustering analysis was pre-set for four clusters, which allowed studying potential direct links between the obtained clusters and the Human Development Index (HDI) categories and the results from the logistic regressions
Summary
We take a look at the multiple dependences presents the correlation plot. The scale of blue between the variables of the models and high- represents positive correlations and the scale light the most correlated variables. Wealth Gini is not significantly correlated with happiness, but income inequalities and poverty lowers it, while HDI (and its underlying pillars of education, health and income) and GDP per capita drive the world happiness. After these suggestive relations, we conduct a clustering analysis and discuss the relatedness of the variables in each constructed cluster of countries. The between_SS / total_SS = 68.3 % gives the total variance in the data explained by the k-means clustering This suggests a relatively good explanatory power of our clustering model, while keeping the number of clusters still relatively low.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
