RPubs link information

• Rpubs link is published.

Introduction

• Nowadays, Suicide attempts are very common. People use different ways for it.
• In US, most of the people uses Gun for suicide than for killing others.
• The place they choose to do suicide varies.

Introduction Cont.

• The analysis based on the Gun death gives an overall idea about:
  -how many people uses gun for suicides and homicides,
  -gender based analysis,
  -what type of place has chosen.
  

Problem Statement

• The main aim of this assignment is to statistically investigate whether there is a relation between gender and the place chosen for suicide or not.
• This will help to take necessary action to reduce the attempts of it.
• The method used to check the relation between sex and suicide committing place is chi-square independence (association) test.

Data

• The dataset used for this assignment is collected from Kaggle.com and it contains information about gun death in the US in the years between 2012-2014.

• The dataset contains data regarding the victim’s age, sex, race, education, intent, time (month and year) and place of death, and weather or not police was at the place of death.

Data Cont.

• The dataset contain 100798 observations and 11 variables.

• The main variables used in this assignment are:

  -Sex: It is the victims gender, and Male and Female is represented by M and F respectively.
  
  
  -Intent: It shows method of death by gun,includes ‘Suicide’, ‘Accidental’, ‘NA’, ‘Homicide’, or
  ‘Undetermined’.
  
  
  -Place: It shows the place that deaths occured like ‘Farm’, ‘Home’, ‘Industrial/construction’, ‘Other specified’, ‘Other unspecified’, ‘Residential institution’, ‘School/institution’, ‘Sports’, ‘Street Trade/service area’.
  

Descriptive Statistics and Visualisation

• First, the dataset is imported and viewd.

guns <- read_csv("guns.csv")
View(guns)

guns$sex <- as.factor(guns$sex)
guns$place <-as.factor(guns$place)

• Then three variables sex, intent and place is selected for doing the analysis.
• It is then filtered to get the data of suicide.
• Then, data is checked to find whether there is a missing value or not.

gun1 <- guns %>% select(sex,place,intent)
gun2 <- gun1 %>%  filter(intent=="Suicide") %>% select(sex,place)
str(gun2)

## Classes 'spec_tbl_df', 'tbl_df', 'tbl' and 'data.frame': 63175 obs. of  2 variables:
##  $ sex  : Factor w/ 2 levels "F","M": 2 1 2 2 2 2 2 2 2 2 ...
##  $ place: Factor w/ 10 levels "Farm","Home",..: 2 9 4 2 4 2 2 2 2 2 ...
##  - attr(*, "spec")=
##   .. cols(
##   ..   X1 = col_double(),
##   ..   year = col_double(),
##   ..   month = col_character(),
##   ..   intent = col_character(),
##   ..   police = col_double(),
##   ..   sex = col_character(),
##   ..   age = col_double(),
##   ..   race = col_character(),
##   ..   hispanic = col_double(),
##   ..   place = col_character(),
##   ..   education = col_double()
##   .. )

Decsriptive Statistics Cont.

• Then, data is checked to find whether there is a missing value or not.

is.na(gun2) %>% sum()

## [1] 0

Decsriptive Statistics Cont.

• The summary shows that females are suicided more than that of males.

gun2[gun2$sex=="M",] %>% summary() #summary of male

##  sex                      place      
##  F:    0   Home              :38691  
##  M:54486   Other specified   : 7182  
##            Other unspecified : 4174  
##            Street            : 1972  
##            Trade/service area: 1518  
##            Farm              :  345  
##            (Other)           :  604

gun2[gun2$sex=="F",] %>% summary() #summary of female

##  sex                     place     
##  F:8689   Home              :6724  
##  M:   0   Other specified   : 849  
##           Other unspecified : 600  
##           Street            : 209  
##           Trade/service area: 205  
##           Farm              :  33  
##           (Other)           :  69

Decsriptive Statistics Cont.

• The detailed view of the places where suicide is committed.

 gun3 <- gun2 %>% table() %>% prop.table(margin = 1) %>% round(4)
gun3

##    place
## sex   Farm   Home Industrial/construction Other specified
##   F 0.0038 0.7739                  0.0014          0.0977
##   M 0.0063 0.7101                  0.0026          0.1318
##    place
## sex Other unspecified Residential institution School/instiution Sports
##   F            0.0691                  0.0018            0.0035 0.0013
##   M            0.0766                  0.0020            0.0049 0.0016
##    place
## sex Street Trade/service area
##   F 0.0241             0.0236
##   M 0.0362             0.0279

Decsriptive Statistics Cont.

• From visualisation, it is clear that the female committing suicide is more in number than male.
• Even though the proportion of place chosen as home is more in both cases, it is slightly high in female case.
• Secondly, other specified areas for suicide are higher in male than female.

gun4<- as.data.frame(gun3)
ggplot(gun4,aes(y=Freq,x=place,fill=place))+geom_bar(stat="identity")+
  facet_grid(~sex)+ylab("proportion")+theme(axis.text.x = element_blank())+
  labs(title = "Distribution of Places for Each Gender")

Decsriptive Statistics Cont.

• You can use the knitr:kable function to print nice HTML tables. Here is an example R code:

gun4 %>% group_by(sex) %>% summarise(Min = min(Freq,na.rm = TRUE),
                                           Q1 = quantile(Freq,probs = .25,na.rm = TRUE),
                                           Median = median(Freq, na.rm = TRUE),
                                           Q3 = quantile(Freq,probs = .75,na.rm = TRUE),
                                           Max = max(Freq,na.rm = TRUE),
                                           Mean = mean(Freq, na.rm = TRUE),
                                           SD = sd(Freq, na.rm = TRUE),
                                           n = n(),
                                           Missing = sum(is.na(Freq))) -> table1
knitr::kable(table1)

Hypothesis Testing

• From the plot, it shows that the men is more likely to opt place other than home to commit suicide when combared to female. Here, a chi-square test is used to check if this is by chance or there is relation between gender and the area selection.

\[H_0: \:shows \ no \ relation \ between \ sex \ and \ place \ opted \ to \ suicide. \] \[H_A: \:shows \ relation \ between \ sex \ and \ place \ opted \ to \ suicide. \]

Hypthesis Testing Cont.

• The expected value is found and it shows each cells count is higher than 5.
• So the assumption is met and test can be continued.

gun_test <- chisq.test(table(gun2$sex,gun2$place))
gun_test$expected

##    
##          Farm      Home Industrial/construction Other specified
##   F  51.98958  6246.315                21.31848        1104.572
##   M 326.01042 39168.685               133.68152        6926.428
##    
##     Other unspecified Residential institution School/instiution   Sports
##   F          656.6092                17.19232          40.57388 13.47878
##   M         4117.3908               107.80768         254.42612 84.52122
##    
##        Street Trade/service area
##   F  299.9717            236.979
##   M 1881.0283           1486.021

Hypthesis Testing Cont.

• From the following observations:
  *it is found that X-squared value is 170.16 and degree of freedom = (2-1)(10-1)=9
• critical value = 16.91898. It is far less than the X-squared value.
• p-value < 0.001.
  *Thus, there is a statistical significant association.

gun_test

## 
##  Pearson's Chi-squared test
## 
## data:  table(gun2$sex, gun2$place)
## X-squared = 170.16, df = 9, p-value < 2.2e-16

qchisq(p=.95 , df=9)

## [1] 16.91898

pchisq(q=170.16,df=9, lower.tail = FALSE)

## [1] 5.714766e-32

Discussion

• The hypothesis test shows that χ2 value is 170.16 and p-value is less than 0.001. This shows that there is a statistically significant association between sex and place. The output suggest that female is more likely to commit suicide at home.

• Limitation: The ‘Other specified’ and ‘Other unspecified’ is included in place category . This has to be more detailed to enhance the accuracy of the result.

• Strength: The sample size is large and the data is collected from different sources. This increases the objectivity of the data.

• Future direction: The results is not enough to determine why female is more likely to attempt suicide at home. This might be the biology feature, but evidence is not clear. There should be more factors need to be involved, such as family status, if they have kids or family violence, etc.

References

• Data set: https://www.kaggle.com/hakabuk/gun-deaths-in-the-us/download

• Orignal data: https://fivethirtyeight.com/features/gun-deaths/