Analytics

Overview of Clustering and Other ML Methods Used

Clustering

To analyze death rate(s) in a meaningful manner we chose to cluster our data using k-means clustering, although during Phase 1 Divisive Analysis (DIANA) Clustering was also shown to be somewhat effective for some parts of the data; R provides clustering functions for both k-means and DIANA. K-means basically tries to group data points into clusters that are the shortest distance from the other points in the cluster. Diana clusters by starting out in one giant cluster and then attempts to partition the data. For the mortality data of interest to MortalityMinder, both approaches produce similar numbers of clusters; k-means was selected because it is easier to add to clusters after we impute the data than with diana. For factors data we analyzed before we imputed. This is because the factors are used to impute the data and the imputed data is not reliable for analysis.

Handling Small States

If a state has six or less counties (esp. Delaware, Hawaii, and Rhode Island), then it was not clustered at all and becomes an exception case. Exception were handled separately in parts of the app. For example, the line chart shows the mortality rate by county and not the mean mortality rate for cluster (both are over several years). The factor distribution maps also do not exist because there are no clusters for us to map. The box-plots become a single line (there is only one value per social determinant per county, so this is mathematically correct) and the dot plot below it does not show clusters (the clusters do not exist).

Correlation Analysis

We used a Kendall Tau test for determining non-parametric rank correlations. Kendall Tau tests whether there is an association between data and returns values from -1 to 1. In this case, 1 means the variable is associated with higher mortality rates and -1 means it is associated with lower mortality rates (or rather the clusters for higher and lower mortality rate respectively). The higher the absolute value, the greater the association. In MortalityMinder we want to discover what medical, economic, and social factors are related to our clusters with higher mortality rates , or counties, in the case of the exception states. Kendall Tau was chosen because it was a familiar, flexible method for analyzing non-parametric rank correlations yielding good results.