Showing posts with label guns. Show all posts
Showing posts with label guns. Show all posts

Thursday, August 8, 2019

When Looking at Gun Ownership and Gun Deaths, Dates Matter



In response to the horrific shootings in El Paso, TX and Dayton, OH, I saw the graph below from Vox showing the correlation between gun ownership and gun deaths.  The graph was used on Late Night with Seth Meyers on Monday right after the shootings in the above clip.  It shows that the US is an extreme outlier in gun deaths and ownership comparred to other developed nations.
In response to the graph I was going to look at what type of a relationship still exists between guns and gun deaths after the US is excluded.  I went to the source for the data for this chart gunpolicy.org to find the data that was used in the its creation.  

The United States had 120.5 guns per 100 individuals and 12.21 total gun deaths per 100,000 individuals both in 2017.  On the graph the US is seen closer to 100 guns per 100 and below 12 deaths/100,000.

The United Kingdom or UK is in the lower left hand corner of the chart near the origin of both axes of the graphs.  The Gun Policy website for the UK lists 5.03 guns/100 for 2017 and 0.2 deaths/100,000 for the year 2015.  

The nation of Cyprus had 34 guns/100 for 2017 (the only time point listed) and 1 death/100,000 for 2016.  On the graph Cyprus is listed closer to 2 deaths/100,000 and closer to 40 guns/100.  

To do a proper correlational study, the data points for each county for each axis need to be from the same time point (year in this case).  There could be different ownership and death rates for the years not listed.  The investigator could be accused of cherry picking the data that supports the investigator's argument.  

I do agree with the argument that more guns lead to more gun deaths but to conduct a proper analysis of this data the methodology must be sound to protect against claims of dishonesty.  For me to do a proper study I will have to sift through the gun policy website and use only gun ownership and death data from a recent year for the countries from which both variables are available for that year.


**Related Posts**

Season's Shootings

Saturday, November 3, 2018

Are Mass Shootings Rare in Southwestern Pennsylvania?



In the press conference following the horrific shootings at the Tree of Life Synagogue which climed 11 lives, Wendell Hissrich, Pittsburgh's director of public safety, said "these events usually occur in other cities, today the nightmare has hit home in the city of Pittsburgh."  The video clip above details six other mass shooting events (defined as events with three or more fatailities) in and around the city that have occurred sine 2000.  These events include Ronald Taylor and Richard Baumhammers racially motivated shooting sprees, Richard Poplawski's killing of three Pittsburgh Police officers a mile from where I lived in Stanton Heights, George Sodini who killed three women at the LA Fitness Center in the South Hills, a drug related shooting in Wilkinsburg and a car wash shooting in Fayette County.



To test this claim I can compare the number of shootings here in Pittsburgh for time and population to the national rate of mass shootings since 2000 adjusting for population and time.  The Poisson distribution is a probability distribution used to estimate the probability of rare events,  In the above graphic x is the number of events of interest, lambda is the mean # of occurrences say the national rate of shootings and x! (called x factorial) is a shorthand way of writing 5! = 5*4*3*2*1 = 120.

Mother Jones magazine maintains an online database of mass shootings related to mass shooters whose motives were not gang related and did not occur in a home.  They list 75 such incidents in their database since 2000.  The shootings by Baumhammers, Sodini, and Taylor were not included but should have been by their criteria so I will include them in the total number of incidents making it 78.  The shooting by Poplawski is not included because it occurred in his home and the other which happened in Wilkinsburg is also not included because it is drug/gang related.  Dividing by the US population in the 2010 census of 308,745,538.  Population adjusted rate of these mass shootings in 0.25 shootings per million residents for lambda in the above equation.  Entering the 5 incidents into the formula for x gives a probability value approaching zero for observing this rate given the national rate.  

If one adjusts the observed incidents for the population of the three counties in which they occurred (1,557,774 from the 2010 census) we find a rate of 3.19 incidents per million which is considerably higher than the national rate 0.25 incidents/million.  If we enter this number into the formula (rounded down to three to make it an integer) we find a probability value of 0.00203 for having exactly three mass shootings here over the period 2000-2018.  The cumulative probability of observing 3 or more shooting per million given the national rate is 0.00216.


Mother Jones Map for Mass Shootings for 1982-2018. Only two incidents from SW PA were included.
If we consider the number of fatalities in last Saturdays shooting, Hissrich is right that the attacks represent a new level of violence here similar to Parkland, or Columbine.  Looking at the number of past mass shootings in the context of the population it looks as though they are more common here than in other parts of the country (though it is not the only hotspot for them as can be seen in the map above).  If drug/gang and domestic violence related shootings are included the numbers would look more horrific for local and the national map.

**Related Posts**

Season's Shootings



Sunday, July 12, 2015

Guns and Hate Groups

Much has been made of Dylann Roof purchasing a gun due to a failed FBI background check before his massacre of 9 African Americans in Charleston, SC.  Mother Jones reported on a study of gun deaths and gun ownership in each state in 2013 by Kalesan et al. in the journal Injury Prevention.

Having looked at the concentration of hate groups in each state (adjusted for the state's population),  I thought I would take a look at how gun ownership relates to the concentration of hate groups in the US.  I was unable to access Kalesan's data but I was able to access data on gun deaths which they showed is correlated with gun ownership.


I was able to access data on gun death rates in the 50 states plus the District of Columbia (DC) in 2013 and I did correlate that data with the rates of hate groups tracked by the Southern Poverty law center.  The correlation coefficient for gun death rates and the concentration of hate groups is 0.52.  DC is an extreme value for both gun deaths and hate groups.  Rerunning the correlation without DC the value becomes 0.47.  A correlation of +1 would produce  a chart where all of the states form a perfect straight line sloping upward.
Plot of State Hate Group rates vs.Gun Death Rates
The chart at the bottom is on that I created in the stat package R (which I am learning).  The point in the upper right hand corner is DC.  It graphically demonstrates the relationship between the concentration of hate groups and the murder rate.  This means that in states with a higher number of hate groups per million, a higher hate of gun deaths is seen.  A cause and effect relationship between these two phenomenon is harder to establish. 

STATE
RATE Gun Deaths
pop2012
Hate groups
Hate Groups per million
AL
17.6
4822023
22
4.56
AK
19.8
731449
1
1.37
AZ
14.1
6553255
20
3.05
AR
16.8
2949131
24
8.14
CA
7.7
38041430
77
2.02
CO
11.5
5187582
17
3.28
CT
4.4
3590347
5
1.39
DE
10.3
917092
4
4.36
DC
22
632323
15
23.72
FL
11.9
19317568
58
3
GA
12.6
9919945
50
5.04
HI
2.6
1392313
0
0
ID
14.1
1595728
9
5.64
IL
8.6
12875255
23
1.79
IN
13
6537334
26
3.98
IA
8
3074186
5
1.63
KS
11.4
2885905
5
1.73
KY
13.7
4380415
15
3.42
LA
19.3
4601893
20
4.35
ME
10.9
1329192
2
1.5
MD
9.7
5884563
15
2.55
MA
3.1
6646144
12
1.81
MI
12
9883360
18
1.82
MN
7.6
5379139
8
1.49
MS
17.8
2984926
22
7.37
MO
14.4
6021988
23
3.82
MT
16.7
1005141
8
7.96
NE
9
1855525
9
4.85
NV
13.8
2758931
8
2.9
NH
6.4
1320718
7
5.3
NJ
5.7
8864590
44
4.96
NM
15.5
2085538
6
2.88
NY
4.2
19570261
42
2.15
NC
12.1
9752073
33
3.38
ND
11.8
699628
1
1.43
OH
11
11544225
31
2.69
OK
16.5
3814820
17
4.46
OR
11
3899353
9
2.31
PA
11.2
12763536
41
3.21
RI
5.3
1050292
3
2.86
SC
15.2
4723723
20
4.23
SD
10
833354
3
3.6
TN
15.4
6456243
37
5.73
TX
10.6
26059203
57
2.19
UT
12.6
2855287
6
2.1
VT
9.2
626011
4
6.39
VA
10.2
8185867
26
3.18
WA
8.7
6897012
10
1.45
WV
14.3
1855413
10
5.39
WI
9.7
5726398
10
1.75
WY
16.7
576412
2
3.47
 


**Related Posts** 

A Wave of Hate Groups in California? No in Washington, DC

A Statistical Profile of the Uninsured in Washington, DC, New Mexico, and Texas

Correlation with the Number of Hate Groups per Million, Poor Health Suggests More Hate