Happiness-Data-EDA

EDA on UN World Happiness Report and World Bank Metrics from 2019

Project Notes

This EDA was performed as part of a group project for a course on Probability and Statistics as part of a Masters of Data Science at Denver University. It was a fully collaborative effort with classmates Don Smith and Brittany Laurent, and results and approaches were reviewed regularly amongst ourselves and our class supervisor.

My Role

My primary contributions to this project were code implementation of Shapiro-Wilk testing, Q-Q Plots, Levene Testing, and Principal Component Analysis, as well as final decisions on the communication of results and data.

File Glossary:

happy_data_augment.csv:

Country-indexed data from 2019 UN World Happiness Report, augmented with 2019 World Bank Metrics

gh_happiness_data_EDA.rmd, .pdf:

R-markdown file with code and explanation of entire EDA process

PCAbiplot.png:

Separate file of "most descriptive" interrelation between predictive variables in our data

Data Glossary

• "country_name"

• "country_code"

• "region"

• “happy_rank” - Ranking of happiness score

• “happy_score” - Aggregate happiness score calculated from all other factors

• “happy_gdpc” - GDP per capita

• “happy_supp” - Sense of social support

• “happy_health” - Healthy life expectancy at birth

• “happy_free” - Happiness with level of personal freedom

• “happy_gen” - How often people contribute to charitable causes

• “happy_trust” - Trust level that own national government is not corrupt

• “wb_pov” - % of population below UN international poverty rate

• “wb_unemp” - % of able-bodied labor force unemployed

• “wb_elec” - % of population with access to electricity

• “wb_renew” - % of final energy use from renewable sources

• “wb_hom” - Homicide rate per 100,000 people

• “wb_debt” - National debt as % of government GDP

• "gdpc_change" - Synthetic variable constructed from comparison of UN World Happiness Report 2018 and 2019 GPDC data

Glossary of Exploratory Approaches

• Studentized Breusch-Pagan Test for Heteroscedasticity of predictor variables
• Shapiro-Wilk Test and Q-Q Plotting for Normality of predictor variables
• Decile-level value imputation
• Pairwise correlation matrix against 'happy_score'
• Shapiro-Wilk Test and Levene Test of 'happy_score'for Normality and Homogeneity of Variance against 'region' info
• One-Way ANOVA of 'happy_score' against 'region'
• Tukey Honest-Significant-Difference Test on all ANOVA results
• Robust Regression Testing via Maximum Likelihood-type M-Estimation (chosen due to conditions of Ordinary-Least-Squares regression not being met)
• Principal Component Analysis (PCA)