This data come from the U.S. National Oceanic and Atmospheric Administration's (NOAA) storm database. The objective was to analyze the effects of different types of natural disasters over the outcomes (fatalities and injuries) related to the people of the U.S.
This database is stored as a comma-separated-value file compressed via the bzip2 algorithm to reduce its size. You can download the file from this link: Storm Data 47Mb
You can read the file by extracting into you working directory and using the function read.csv. I strongly recommend that you format the column BGN_DATE with the functions as.POSIXlt for any further investigation related to month/year of this data. However it is not needed for this short analysis. The graphics of this work are based on the ggplot2 codes. So for proper working use the library function to load ggplot2 before the analysis.
StormData <- read.csv("repdata-data-StormData.csv")
df <- within(StormData, datetime <- as.POSIXlt(BGN_DATE, format = "%m/%d/%Y"))
library(ggplot2)
require(gridExtra)
## Loading required package: gridExtra
## Loading required package: grid
A few tables ware constructed. The first table, called type, was intend to check the number of different disasters and its frequency. This same result can be seen by the use of the function summary, however, this table will be used for later propose. With the use of this last function, we can also see the mean fatalities and injuries, and the most common state to have natural disasters. The use of the function str shows us that there are 985 different natural disasters.
type <- data.frame(table(StormData$EVTYPE))
summary(df[, c(7, 8, 23, 24)], 10)
## STATE EVTYPE FATALITIES
## TX : 83728 HAIL :288661 Min. : 0
## KS : 53440 TSTM WIND :219940 1st Qu.: 0
## OK : 46802 THUNDERSTORM WIND : 82563 Median : 0
## MO : 35648 TORNADO : 60652 Mean : 0
## IA : 31069 FLASH FLOOD : 54277 3rd Qu.: 0
## NE : 30271 FLOOD : 25326 Max. :583
## IL : 28488 THUNDERSTORM WINDS: 20843
## AR : 27102 HIGH WIND : 20212
## NC : 25351 LIGHTNING : 15754
## (Other):540398 (Other) :114069
## INJURIES
## Min. : 0.0
## 1st Qu.: 0.0
## Median : 0.0
## Mean : 0.2
## 3rd Qu.: 0.0
## Max. :1700.0
##
##
##
##
str(type)
## 'data.frame': 985 obs. of 2 variables:
## $ Var1: Factor w/ 985 levels " HIGH SURF ADVISORY",..: 1 2 3 4 5 6 7 8 9 10 ...
## $ Freq: int 1 1 1 1 4 1 1 1 1 4 ...
Other tables were made to sum the number of injuries or deaths for each type of disasters. Though the column bind function, these sums were attach to the “type” table. From this combined table it was possible to check the mean values of fatalities and injuries for each type of disaster. These tables were ordered by most frequent. A table of the separated number of outcomes order by the most “chaotic” was also created.
dffatsum <- aggregate(FATALITIES ~ EVTYPE, data = df, sum, na.action = na.omit)
dfinjsum <- aggregate(INJURIES ~ EVTYPE, data = df, sum, na.action = na.omit)
dffs <- cbind(type, dffatsum$FATALITIES, dfinjsum$INJURIES)
colnames(dffs) <- c("EVTYPE", "FREQ", "FATALITIES", "INJURIES")
dffs$FATMEAN <- dffs$FATALITIES/dffs$FREQ
dffs$INJMEAN <- dffs$INJURIES/dffs$FREQ
FATALITIES <- head(dffs[order(dffs$FATALITIES, decreasing = TRUE), ], 20)[,
c(1, 3)]
INJURIES <- head(dffs[order(dffs$INJURIES, decreasing = TRUE), ], 20)[, c(1,
4)]
FATMEAN <- head(dffs[order(dffs$FATMEAN, decreasing = TRUE), ], 20)[, c(1, 5)]
INJMEAN <- head(dffs[order(dffs$INJMEAN, decreasing = TRUE), ], 20)[, c(1, 6)]
dffat <- head(df[order(df$FATALITIES, decreasing = TRUE), ], 40)[, c(8, 23)]
dfinj <- head(df[order(df$INJURIES, decreasing = TRUE), ], 70)[, c(8, 24)]
For exception of tornados, it is possible to see that the most 10 frequent catastrophe are not as well the most responsible for the highest number of outcomes related to people. Bellow it's possible to show the 10 most frequent disasters:
head(dffs[order(dffs$FREQ, decreasing = TRUE), ], 10)
## EVTYPE FREQ FATALITIES INJURIES FATMEAN INJMEAN
## 244 HAIL 288661 15 1361 5.196e-05 0.004715
## 856 TSTM WIND 219940 504 6957 2.292e-03 0.031631
## 760 THUNDERSTORM WIND 82563 133 1488 1.611e-03 0.018023
## 834 TORNADO 60652 5633 91346 9.287e-02 1.506067
## 153 FLASH FLOOD 54277 978 1777 1.802e-02 0.032739
## 170 FLOOD 25326 470 6789 1.856e-02 0.268064
## 786 THUNDERSTORM WINDS 20843 64 908 3.071e-03 0.043564
## 359 HIGH WIND 20212 248 1137 1.227e-02 0.056254
## 464 LIGHTNING 15754 816 5230 5.180e-02 0.331979
## 310 HEAVY SNOW 15708 127 1021 8.085e-03 0.064999
In addition, due to the relationship of frequency of events, the disasters with the highest number of incidents doesn't have the highest mean (outcome/ disaster frequency). These numbers shows us that the average number of deaths and injuries doesn't represent a valid analysis, since the mean is highest to rare events. However, this brings the knowledge that people are less prepared to these events, therefore, they bring higher risks of incidents.
head(dffs[order(dffs$FATMEAN, decreasing = TRUE), ], 10)
## EVTYPE FREQ FATALITIES INJURIES FATMEAN INJMEAN
## 842 TORNADOES, TSTM WIND, HAIL 1 25 0 25.000 0.000
## 72 COLD AND SNOW 1 14 0 14.000 0.000
## 851 TROPICAL STORM GORDON 1 8 43 8.000 43.000
## 580 RECORD/EXCESSIVE HEAT 3 17 0 5.667 0.000
## 142 EXTREME HEAT 22 96 155 4.364 7.045
## 279 HEAT WAVE DROUGHT 1 4 15 4.000 15.000
## 373 HIGH WIND/SEAS 1 4 0 4.000 0.000
## 487 MARINE MISHAP 2 7 5 3.500 2.500
## 976 WINTER STORMS 3 10 17 3.333 5.667
## 340 Heavy surf and wind 1 3 0 3.000 0.000
head(dffs[order(dffs$INJMEAN, decreasing = TRUE), ], 10)
## EVTYPE FREQ FATALITIES INJURIES FATMEAN INJMEAN
## 277 Heat Wave 1 0 70 0.0000 70.00
## 851 TROPICAL STORM GORDON 1 8 43 8.0000 43.00
## 954 WILD FIRES 4 3 150 0.7500 37.50
## 821 THUNDERSTORMW 1 0 27 0.0000 27.00
## 366 HIGH WIND AND SEAS 1 3 20 3.0000 20.00
## 656 SNOW/HIGH WINDS 2 0 36 0.0000 18.00
## 224 GLAZE/ICE STORM 1 0 15 0.0000 15.00
## 279 HEAT WAVE DROUGHT 1 4 15 4.0000 15.00
## 973 WINTER STORM HIGH WINDS 1 1 15 1.0000 15.00
## 411 HURRICANE/TYPHOON 88 64 1275 0.7273 14.49
A graphical interpretation of the 20 events with highest average:
plot3 <- ggplot(FATMEAN, aes(as.character(EVTYPE), FATMEAN)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = "Top 20 disasters with most number of deaths per event",
x = NULL, y = "Number of Fatalities")
plot4 <- ggplot(INJMEAN, aes(as.character(EVTYPE), INJMEAN)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = "Top 20 disasters with most number of injuries per event",
x = NULL, y = "Number of Injuries")
grid.arrange(plot3, plot4, ncol = 2)
Furthermore, to better interpretation of this data, an analysis of the sum of outcomes can be better analyzed with a graphical view. This graphic shows that tornados are the biggest responsible for injuries and fatalities even though its average is low.
plot1 <- ggplot(FATALITIES, aes(as.character(EVTYPE), FATALITIES)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = "Top 20 disasters with most number of deaths",
x = NULL, y = "Number of Fatalities")
plot2 <- ggplot(INJURIES, aes(as.character(EVTYPE), INJURIES)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = "Top 20 disasters with most number of injuries",
x = NULL, y = "Number of Injuries")
grid.arrange(plot1, plot2, ncol = 2)
subset(dffs, dffs$EVTYPE == "TORNADO")
## EVTYPE FREQ FATALITIES INJURIES FATMEAN INJMEAN
## 834 TORNADO 60652 5633 91346 0.09287 1.506
In the next graphic, it's possible to show the 7 worst and most devastating event of each disasters. By this graphic we can see that, even though tornados are the most frequent cause of deaths and injuries overall, excessive heat, extreme heat and heat, when happens, have the highest impact in numbers of fatalities.
plot5 <- ggplot(dffat, aes(as.character(EVTYPE), FATALITIES)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = NULL,
x = NULL, y = "Number of Fatalities")
plot6 <- ggplot(dfinj, aes(as.character(EVTYPE), INJURIES)) + geom_bar(stat = "identity",
position = "dodge") + theme(axis.text.x = element_text(angle = 90)) + labs(title = NULL,
x = NULL, y = "Number of Injuries")
grid.arrange(plot5, plot6, ncol = 2, main = "Most devastating disater for the worst 7 types of events")
