The following report analyzes data form the U.S. National Oceanic and Atmospheric Administration’s (NOAA) storm database. The database contains information about weather events in the USA between 1950 and 2011. The goal of this analysis is to determine which of those events represents a major threat to the US population from both a health and economic perspective. Three particular variables are then considered for this report: Fatalities, Injuries, and Property Damage
Data set is loaded and three subsets are created. Property Damage in Dollars, number of injuries and number of deaths, per event type are each contained in a different data subset.
##Packages needed
library(ggplot2)
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(ggpubr)
NOOA<-read.csv("repdata_data_StormData.csv")
Injuries<-aggregate(INJURIES ~ EVTYPE, data=NOOA, sum)
property<-aggregate(PROPDMG ~ EVTYPE,data=NOOA,sum)
Fatalities<-aggregate(FATALITIES ~ EVTYPE, data=NOOA, FUN=sum)
Each data subset is ranked in decreasing order, and the 10 first entry are stored in a new individual dataset. Through this process it is possible to identify the whether events that usually generate more economic and health damage.
FatalitiesOrdered<-Fatalities[order(Fatalities$FATALITIES,decreasing = T),]
FatalitiesTop10<-FatalitiesOrdered[1:10,]
head(FatalitiesTop10)
## EVTYPE FATALITIES
## 834 TORNADO 5633
## 130 EXCESSIVE HEAT 1903
## 153 FLASH FLOOD 978
## 275 HEAT 937
## 464 LIGHTNING 816
## 856 TSTM WIND 504
Injuriesordered<-Injuries[order
(Injuries$INJURIES,decreasing = TRUE),]
InjuriesTop10<- Injuriesordered[1:10,]
head(InjuriesTop10)
## EVTYPE INJURIES
## 834 TORNADO 91346
## 856 TSTM WIND 6957
## 170 FLOOD 6789
## 130 EXCESSIVE HEAT 6525
## 464 LIGHTNING 5230
## 275 HEAT 2100
Before the Property damage top 10 is created, it is first necessary to recode the PROPDMGEXP (a column containing the exponential indicator) into appropriate ‘multipliers’. The letters in that column are transformed into multipliers. That way the real value of the damage in dollars can be calculated by multiplying the PROPDMG and the PROPDMGEXP columns.
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "H"] <- 100 #H-> Hundreds
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "h"] <- 100
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "K"] <- 1000 #K-> Thousands
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "M"] <- 1e+06 #M-> Millions
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "m"] <- 1e+06
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "B"] <- 1e+09 #B-> Billions
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == ""] <- 1
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "0"] <- 1
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "1"] <- 10
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "2"] <- 100
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "3"] <- 1000
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "4"] <- 10000
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "5"] <- 1e+05
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "6"] <- 1e+06
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "7"] <- 1e+07
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "8"] <- 1e+08
# Invalid values
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "+"] <- 0
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "-"] <- 0
NOOA$PROPDMGEXP[NOOA$PROPDMGEXP == "?"] <- 0
#New column with total amount of property damage in dollars
NOOA$PROPDMGEXP<-as.numeric(NOOA$PROPDMGEXP)
NOOA$TOTALPROPDMG<-NOOA$PROPDMG * NOOA$PROPDMGEXP
The 10 most harmful events in terms of property damage can now be calculated and stored in a new dataset.
Damage<-aggregate(TOTALPROPDMG~EVTYPE,data=NOOA,FUN=sum)
Damageordered<-Damage[order(Damage$TOTALPROPDMG,decreasing = TRUE),]
Damagetop10<-Damageordered[1:10,]
head(Damagetop10)
## EVTYPE TOTALPROPDMG
## 170 FLOOD 144657709870
## 411 HURRICANE/TYPHOON 69305840000
## 834 TORNADO 56947381845
## 670 STORM SURGE 43323536000
## 153 FLASH FLOOD 16822678195
## 244 HAIL 15735270147
Now that the three groups of top 10 weather events have been created it would be convenient to identify which of those events appear in the three rankings, which appear in at least two of them, and which appear only in one. This should give use some clue about the events that are more harmful from an economic and health perspective. That information (# of appearances in the rankings) can then be plotted using colors.
Three groups of names are created.
Group 1 contains the names of all the weather events in the rankings.
Group 2 contains the names of the events that appear two times in the rankings.
Group 3 contains the names that appear in the three rankings.
i.e (FatalitiesTop10,InjuriesTop10,Damagetop10)
##Complements
#1 All names in the ranking (without duplicates)
`%notin%` <- Negate(`%in%`)
Fat<-FatalitiesTop10$EVTYPE
Inj<-InjuriesTop10$EVTYPE
Dam<-Damagetop10$EVTYPE
all_names<- c(Fat,Inj,Dam)
#Names of whether events that appear in at least two Rankings
Fat_Inj<-which(Fat %in% Inj)
names_Fat_Inj<- FatalitiesTop10[Fat_Inj,1]
Fat_Dam<-which(Fat%in%Dam)
names_Fat_Dam<-FatalitiesTop10[Fat_Dam,1]
In_at_least_two_rankings<-unique(c(names_Fat_Inj,names_Fat_Dam))
#Group 3 Names of Weather events that appear in all three rankings
Fat_Inj_Dam<- which(names_Fat_Inj%in%Dam)
names_Fat_InJ_Dam<- names_Fat_Inj[Fat_Inj_Dam]
Names_Group3<-names_Fat_InJ_Dam
##Group 2 Events that appear in only in two Rankings.
Group2<- which(In_at_least_two_rankings%notin%names_Fat_InJ_Dam)
Names_Group2<- In_at_least_two_rankings[which(In_at_least_two_rankings%notin%names_Fat_InJ_Dam)]
#3 Group 1 Events that appear only in one Ranking
Unique_all_names<- unique(all_names)
Group1<- which(Unique_all_names%notin%In_at_least_two_rankings)
Names_Group1<- Unique_all_names[which(Unique_all_names%notin%In_at_least_two_rankings)]
View(Names_Group1)
View(Names_Group2)
View(Names_Group3)
A for loop can now be used to create columns in each ranking that indicate the group at which each event belongs (Group 1, Group 2, Group 3). When creating the plots (with ggplot), this variable can be passed as an argument of “fill”.
A<-nrow(FatalitiesTop10)
for (i in 1:A) {
if (FatalitiesTop10[i,1]%in%Names_Group1) {
FatalitiesTop10[i,3] <- "Group1"}
else if (FatalitiesTop10[i,1]%in%Names_Group2){
FatalitiesTop10[i,3] <- "Group2"}
else if (FatalitiesTop10[i,1]%in%Names_Group3){
FatalitiesTop10[i,3] <- "Group3"}
}
FatalitiesTop10<- FatalitiesTop10%>%rename("Group"="V3")
head(FatalitiesTop10)
## EVTYPE FATALITIES Group
## 834 TORNADO 5633 Group3
## 130 EXCESSIVE HEAT 1903 Group2
## 153 FLASH FLOOD 978 Group3
## 275 HEAT 937 Group2
## 464 LIGHTNING 816 Group2
## 856 TSTM WIND 504 Group2
B<-nrow(Damagetop10)
for (i in 1:B) {
if (Damagetop10[i,1]%in%Names_Group1) {
Damagetop10[i,3] <- "Group1"}
else if (Damagetop10[i,1]%in%Names_Group2){
Damagetop10[i,3] <- "Group2"}
else if (Damagetop10[i,1]%in%Names_Group3){
Damagetop10[i,3] <- "Group3"}
}
Damagetop10<- Damagetop10 %>% rename("Group" = "V3")
head(Damagetop10)
## EVTYPE TOTALPROPDMG Group
## 170 FLOOD 144657709870 Group3
## 411 HURRICANE/TYPHOON 69305840000 Group1
## 834 TORNADO 56947381845 Group3
## 670 STORM SURGE 43323536000 Group1
## 153 FLASH FLOOD 16822678195 Group3
## 244 HAIL 15735270147 Group1
C<-nrow(InjuriesTop10)
for (i in 1:C) {
if (InjuriesTop10[i,1]%in%Names_Group1) {
InjuriesTop10[i,3] <- "Group1"}
else if (InjuriesTop10[i,1]%in%Names_Group2){
InjuriesTop10[i,3] <- "Group2"}
else if (InjuriesTop10[i,1]%in%Names_Group3){
InjuriesTop10[i,3] <- "Group3"}
}
InjuriesTop10<- InjuriesTop10 %>%rename("Group" = "V3")
head(InjuriesTop10)
## EVTYPE INJURIES Group
## 834 TORNADO 91346 Group3
## 856 TSTM WIND 6957 Group2
## 170 FLOOD 6789 Group3
## 130 EXCESSIVE HEAT 6525 Group2
## 464 LIGHTNING 5230 Group2
## 275 HEAT 2100 Group2
With the new columns in each data set, plots can now be created.
Plot1 <- ggplot(data=FatalitiesTop10,
aes(x=reorder(EVTYPE, FATALITIES), y=FATALITIES,fill=Group)) +
geom_bar(stat="identity") + coord_flip() +
scale_fill_manual(values=c("#999999", "#E69F00", "#56B4E9"))+
labs(title="Top 10 fatality causing weather events - USA",
x= "Event Type",
y="Number of Fatalities")
Plot1
Plot2<-ggplot(InjuriesTop10,aes(x= reorder(EVTYPE,INJURIES),y=INJURIES,fill=Group))+
geom_bar(stat = "identity")+coord_flip()+
scale_fill_manual(values=c("#999999", "#E69F00", "#56B4E9"))+
labs(title="Top 10 injuries causing weather events - USA",
x= "Event Type",
y="Number of Injuries")
Plot2
Plot3<-ggplot(Damagetop10,aes(x=reorder(EVTYPE,TOTALPROPDMG),TOTALPROPDMG, fill=Group))+
geom_bar(stat="identity")+
coord_flip()+
scale_fill_manual(values=c("#999999","#E69F00","#56B4E9"))+
labs(title="Top 10 weather events with biggest Property Damage - USA",
x= "Event Type",
y="Damage in Dollars")
Plot3
From the results on the previous plots, it can be said that the three most harmful weather events from both a health and economic perspective are:
-Tornado
-Flood
-Flash Flood