Aim

What are some of the factors that contribute gun violence, and what is the soical media response made by politicians in light of these crisis?

There have a total of 116 Mass Shootings since 1982, where the number of fatalities per shooting has been three or more. Where five of the most deadliest shooting have occured since 2007. There have been two independent studies conducted by Harvard University and the Federal Bureau of Investigation (FBI), where they have found not only has there been an increas of incidents but also they have proven that the deadliness of the shootings have also been increasing in recent times. Although these incidents seem isolated (only affecting the those directly involved), it infact affects the USA as a whole. According to a report published by the vice chair of the Joint Economic Committee on September of 2019, they have found that in this year alone, America has spent a total of $229 Billion Dollars.

For this project, we decided to focus on three destinct areas:

• Demographics - Are mass shootings correlated with demographic factors, e.g. average income?
• Legal - Do gun control laws influence the rates and/or severity of mass shooting events?
• Social - What is the reaction on social media during/around the time of mass shootings by politicians?

To complete this project we utilized several different sources of data. This data included strucutured and unstructured data. Our goal was to find data pretaining to statitics on demographics for each individual state of America as well as data that contained the state's gun laws. To gague the reaction on social media, chose a data set that focused on Tweets made by policitians in the 115th US Congress. The following are some of the main data sets we used, as well as the links to them.

• Youth poverty - percentage of under-18s living below the national poverty line by state, for selected years 1990, 2000, 2010, 2015, and 2017 www.census.gov/data/tables/time-series/demo/income-poverty/historical-poverty-people
• Average income - average household income by state in $, for selected years 1990, 2000, 2005, 2010, 2012, 2013, 2014, 2015, and 2016  www.census.gov/data/tables/time-series/demo/income-poverty/historical-income-households
• Population Density - average number of inhabitants per square mile by state, for selected years 1960, 1970, 1980, 1990, 2000, and 2010  www.census.gov/data/tables/2010/dec/density-data-text
• Religiosity - percentage of "highly religious" adults in terms of believe in God, importance of religion, worship attendance, and frequency of prayer, by state, survey from 2016  www.pewresearch.org
• Mental health expenses - percentage of under-18s living below the national poverty line by state, for selected years 1990, 2000, 2010, 2015, and 2017 www.kff.org
• Twitter Data - contains of 2 million tweets made by politicians of the 115th US Congress  https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/UIVHQR
• Mass Shootings in America - contains information for all mass shootings (where fatalities > 1) from 1982-2019 www.motherjones.com/politics/2012/12/mass-shootings-mother-jones-full-data
• Rand Gun Laws - list of laws pretaining to firearms for each state www.rand.org

Since we used various types of data (statistical, textural, etc.), we were able to apply a variety of data science techniques for this project. We used time series analaysis when analysing data in which we needed to adjust window size, such as with finding tweets that fall within a certain date range of an incident. When it came to analaysis correlations we focused on a state level and was able to employ the use of both Spearman and Pearson l to find correlations between and incident and a state statistic. For analayzing textual data, we untilized td-idf in addition to suggested stopwords to find meaning text within a tweet. Additional, we were able to use bag of words to find the most frequent terms used by politicians before and after an incident. Lastly, due to its complexity, we used multilevel modeling to gain useful insights on our gun laws data.

We were able to find several different insights on the topic of this project. As part of our research, we have determined the correlation between five different demographic determinants and the absolute number of incidents related to gun violence in schools. Our outcomes comprise heat maps representing the characteristics of all evaluated attributes per state as well as results of the correlation analysis according to Pearson and Spearman. The following visualizations depicts one of those correlations.

When it comes to guns laws, we found that states with higher numbers of gun related incidents tend to have higher number of adoptions of new gun laws throughout the years we studied. This can be viewed in the chart below.

For the reaction on social media by politicians, we focused on tweets made one month prior, one week prior, one day prior, the day of, one day after, one week after, and one month after an incident. We found most frequent words the day after, one week after, and one month after an incident where words relating to gun violence. Constradictory of the post made prior to an event. Given below are the most common words found in the tweets of politicians one day prior, and one day after an event has taken place.