SHAP Value Methods was applied to check the associations between suicide rates and rurality, altitude, climate factors, and social economic factors from 2010 to 2020. Suicide rates are divided to 3 groups: suicide by all means, suicides by firearms, suicides by non-firearms. SHAP values are investigated in each single year and multiple 5-year periods such as 2010 to 2014, 2011 to 2015 ans so on.
Climate Factors includes mean temperature, visibility, precipitation and so on. The code to collect climate data can be found here: Climate Factors
Social economic factors includes poverty, unemployment, disabled, and so on. The code to collect social economic data from Census API can be found here: Social Economic Factors
Since only about 1000 counties have the suicide death records from CDC, I imputed the suicide rates by assigning the average suicide rates from the abjecent counties. The code to impute suicide rates can be found here: Suicide Rate Imputation
Here I am using SHAP Values in year 2016 as an example to show you how I built the model and get the results.
The report for this project can be found here: Developing Interpretable Method to Understand Feature Importance and Interactions for for Suicide Risk Management
