Preparing for US Flu Season
I performed an analysis of data about US flu deaths and population to determine allocation of additional hospital staff.
OVERVIEW:
​
In this project, I analyzed data about US flu deaths and population and performed statistical hypothesis testing to help determine the allocation of additional hospital staff. Using Excel, I cleaned, integrated, and transformed the data before conducting a statistical hypothesis test. Then, I created spatial and temporal data visualizations in Tableau. These results were used to build a presentation of my conclusions and recommendations to assist in determining how to allocate additional medical staff.
​
​
PURPOSE & CONTEXT:
​
-
A hypothetical medical staffing agency wants to proactively plan for staffing needs.
​
-
For a portfolio project, I analyzed data from the CDC and US Census Bureau to examine trends.
​
-
I used the results of my analysis to provide recommendations regarding the allocation of additional hospital staff.
​
DATA:
​
-
US flu deaths by geography, time, age, and gender (source: CDC)
-
US population data by geography (source: US Census Bureau)
TOOLS & TECHNIQUES:
​
-
Excel:
-
Data cleaning, integration, and transformation (including VLOOKUPs)
-
Statistical hypothesis testing
-
-
Tableau:
-
Visual analysis​
-
Storyboard
-
PROCESS:
First, I cleaned the two data sets and created data profiles, addressing data integrity issues and implementing data quality measures. Next, I integrated the two data sets using data transformations. I conducted a statistical analysis, testing for correlation between the variables for percentage of population over 65 and mortality rate. Unfortunately, there was no correlation between the two variables. I went on to perform a statistical hypothesis test and compiled an interim report consolidating my findings (see below).
​
Next, I created spatial, temporal, and statistical data visualizations in Tableau. Finally, I designed a Tableau storyboard presentation communicating my research findings, insights, and recommendations.
STATISTICAL INSIGHTS
One assumption was that vulnerable populations suffer the most from the flu and are most likely to end up in hospitals, which would require additional staff. Adults over 65 make up a considerable part of a state's vulnerable population. Therefore, I chose to focus on the variables for flu mortality rate and percentage of population over 65 in US states.
Initially, I tested for correlation between these two variables. Unfortunately, the correlation coefficient was 0.108, indicating no correlation.
Next, I performed a two-tailed statistical hypothesis test. I decided to test if there was a significant difference in flu mortality rate between a group of states with a higher percentage of its population over 65 and a group of states with a lower percentage of its population over 65. The group of states with a high percentage of population over 65 included Connecticut, Florida, Iowa, Maine, New York, Pennsylvania, Rhode Island, and West Virginia. The group of states with a low percentage of population over 65 included California, Colorado, Georgia, Louisiana, Nevada, Texas, Utah, and Virginia.
I used a significance level of 0.05. With a p-value of 0.0032, I was able to reject my null hypothesis. Therefore, at an alpha of 0.05, there was a significant difference in the average influenza mortality rate between a group of states with a higher percentage of its population over 65 and a group of states with a lower percentage of its population over 65.
Check out my interim report to read about the statistical hypothesis test in greater detail:
DATA VISUALIZATIONS IN FINAL PRESENTATION
To interact with the various visualizations, please use this link to view them through Tableau: https://public.tableau.com/shared/FC6QCP638?:display_count=n&:origin=viz_share_link
RECOMMENDATIONS:
To view my video presentation, please visit this link: https://youtu.be/iyhSm86Wi2U
