HW2 Solutions
Alexander Cardazzi
Quarantine, 2020
Use the sharks dataset for this HW. This should be more slightly more difficult than the first homework. You will likely have to do some googling, but I have hinted some functions to look into. After each line, try to view the data to see if the results of your code match what you wanted it to do.
- Make a time series plot of the number of attacks per year.
- Make a frequency distribution for the ages of those being attacked.
- Plot the locations of the shark attacks. Color the attacks by whether or not the attack was fatal. Once this works, try and get fancy with the
scalespackage. - Make a new binary variable for whether the attack was on the west coast. (hint,
?ifelse) - Make a binary variable for if the attack was unprovoked. (hint,
?ifelse) - Make a binary variable for if there was “No injury” or a “minor” injury (hint, look for these inside the text of the Injury column. hint,
?grepl) - Make a binary variable for if the person attacked was male. You may have to use both the Sex column and the Name column.
- Make summary stats for the sharks data.
- Use the binary variable for no injury or minor injury as your dependent variable. Given an attack, what factors lead to more serious injuries? Run a few regressions, have fun with it.
- Is there a variable that significantly changes the sample size of your regressions? Why
- Do you think these missing ages are missing at random?
- Are serious injuries more or less likely given missingness for the variable?
library("stargazer")##
## Please cite as:## Hlavac, Marek (2018). stargazer: Well-Formatted Regression and Summary Statistics Tables.## R package version 5.2.2. https://CRAN.R-project.org/package=stargazerlibrary("scales")
rm(list = ls())
sharks <- read.csv("C:/Users/alexc/Desktop/Empirical Workshop/data/sharks.csv", stringsAsFactors = FALSE)
#View(sharks)1 - Make a time series plot of the number of attacks per year.
timeSeries <- table(sharks$Year)
plot(timeSeries, type = "l", ylab = "Frequency", xlab = "Year")
2 - Make a frequency distribution for the ages of those being attacked.
plot(table(as.numeric(sharks$Age)), ylab = "Frequncy", xlab = "Age")## Warning in table(as.numeric(sharks$Age)): NAs introduced by coercion
3 - Plot the locations of the shark attacks. Color the attacks by whether or not the attack was fatal.
sharks$fatalColor <- as.numeric(as.factor(sharks$Fatal..Y.N.))
#plot(sharks$Longitude, sharks$Latitude, pch = 19, col = c(1,2))
plot(sharks$Longitude, sharks$Latitude,
col = alpha(sharks$fatalColor, .1), pch = 19)
4 - Make a new binary variable for whether the attack was on the west coast.
sharks$westCoast <- ifelse(sharks$Longitude < -100, 1, 0)
table(sharks$westCoast)##
## 0 1
## 761 2375 - Make a binary variable for if the attack was unprovoked.
table(sharks$Type)##
## Invalid Provoked Questionable
## 61 57 4
## Sea Disaster Under investigation Unprovoked
## 2 1 837
## Watercraft
## 36sharks$unprovoked <- ifelse(sharks$Type == "Unprovoked", 1, 0)
table(sharks$unprovoked)##
## 0 1
## 161 8376 - Make a binary variable for if there was “No injury” or a “minor” injury
sharks$severe_injury <- ifelse(grepl("No injury", sharks$Injury) | grepl("minor", tolower(sharks$Injury)), 0, 1)
table(sharks$severe_injury)##
## 0 1
## 229 7697 - Make a binary variable for if the person attacked was male.
sharks$Sex <- ifelse(sharks$Name == "female", "F", ifelse(sharks$Name == "male", "M", sharks$Sex))
sharks$Sex <- ifelse(sharks$Sex == "", NA, sharks$Sex)
sharks$male <- ifelse(is.na(sharks$Sex), NA, ifelse(sharks$Sex == "M", 1, 0))
table(sharks$male)##
## 0 1
## 224 7558 - Make summary stats for the sharks data.
stargazer(sharks, type = "text")##
## =============================================================================
## Statistic N Mean St. Dev. Min Pctl(25) Pctl(75) Max
## -----------------------------------------------------------------------------
## Year 998 2,009.572 5.588 2,000 2,005 2,014 2,019
## original.order 998 5,297.683 640.417 4,239 4,739.2 5,842 6,427
## Latitude 998 29.611 5.116 5.889 27.355 32.819 60.691
## Longitude 998 -94.624 26.284 -169.534 -96.548 -80.327 -64.621
## Maximum.Temperature 980 82.559 7.805 42.200 79.100 87.900 98.500
## Minimum.Temperature 980 68.324 9.093 23.100 63.750 74.600 84.800
## Temperature 980 75.079 7.986 35.900 71.900 80.500 91.400
## Precipitation 980 0.154 0.653 0.000 0.000 0.000 11.300
## Precipitation.Cover 980 7.362 16.883 0.000 0.000 8.300 100.000
## Wind.Speed 979 14.200 4.556 0.000 11.200 16.400 44.700
## Wind.Gust 126 32.548 7.686 4.500 30.000 36.425 57.500
## Visibility 975 9.233 1.147 1.500 8.800 10.000 12.800
## fatalColor 998 1.964 0.277 1 2 2 3
## westCoast 998 0.237 0.426 0 0 0 1
## unprovoked 998 0.839 0.368 0 1 1 1
## severe_injury 998 0.771 0.421 0 1 1 1
## male 979 0.771 0.420 0.000 1.000 1.000 1.000
## -----------------------------------------------------------------------------9 - Given an attack, what factors lead to more serious injuries?
#exploring weather with these regressions
reg1 <- lm(severe_injury ~ westCoast, data = sharks)
reg2 <- lm(severe_injury ~ westCoast + Temperature + Precipitation + Wind.Speed + Visibility, data = sharks)
reg3 <- lm(severe_injury ~ westCoast*Temperature + Precipitation + Wind.Speed + Visibility, data = sharks)
stargazer(reg1, reg2, reg3,
type = "text", omit.stat = c("ser", "f"))##
## ===================================================
## Dependent variable:
## -----------------------------
## severe_injury
## (1) (2) (3)
## ---------------------------------------------------
## westCoast -0.297*** -0.256*** -0.914***
## (0.030) (0.035) (0.271)
##
## Temperature 0.006*** 0.003
## (0.002) (0.002)
##
## Precipitation -0.002 -0.001
## (0.020) (0.019)
##
## Wind.Speed 0.003 0.002
## (0.003) (0.003)
##
## Visibility 0.002 -0.002
## (0.011) (0.011)
##
## westCoast:Temperature 0.009**
## (0.004)
##
## Constant 0.841*** 0.288* 0.627***
## (0.015) (0.164) (0.214)
##
## ---------------------------------------------------
## Observations 998 975 975
## R2 0.090 0.111 0.116
## Adjusted R2 0.089 0.106 0.111
## ===================================================
## Note: *p<0.1; **p<0.05; ***p<0.01#exploring the person/interaction
reg1 <- lm(severe_injury ~ westCoast*Temperature + unprovoked, data = sharks)
reg2 <- lm(severe_injury ~ westCoast*Temperature + male, data = sharks)
reg3 <- lm(severe_injury ~ westCoast*Temperature + unprovoked + male, data = sharks)
reg4 <- lm(severe_injury ~ westCoast*Temperature + unprovoked + male + as.numeric(Age), data = sharks)## Warning in eval(predvars, data, env): NAs introduced by coercionreg5 <- lm(severe_injury ~ westCoast*Temperature + unprovoked + male + as.numeric(Age) + Year, data = sharks)## Warning in eval(predvars, data, env): NAs introduced by coercionstargazer(reg1, reg2, reg3, reg4, reg5,
add.lines = list(c("Year FE", rep("N", 4), "Y")),
type = "text", omit.stat = c("ser", "f"), omit = "Year")##
## =======================================================================
## Dependent variable:
## -------------------------------------------------
## severe_injury
## (1) (2) (3) (4) (5)
## -----------------------------------------------------------------------
## westCoast -0.972*** -0.998*** -1.021*** -1.050*** -1.052***
## (0.264) (0.263) (0.263) (0.297) (0.294)
##
## Temperature 0.002 0.002 0.002 0.002 0.002
## (0.002) (0.002) (0.002) (0.003) (0.003)
##
## unprovoked 0.078** 0.066* -0.018 -0.028
## (0.035) (0.035) (0.044) (0.043)
##
## male -0.048 -0.046 -0.057* -0.068**
## (0.030) (0.030) (0.033) (0.033)
##
## as.numeric(Age) 0.001 0.001
## (0.001) (0.001)
##
## westCoast:Temperature 0.010*** 0.011*** 0.011*** 0.012*** 0.012***
## (0.004) (0.004) (0.004) (0.004) (0.004)
##
## Constant 0.618*** 0.710*** 0.682*** 0.746*** 20.865***
## (0.185) (0.189) (0.189) (0.211) (4.907)
##
## -----------------------------------------------------------------------
## Year FE N N N N Y
## Observations 980 963 963 761 761
## R2 0.120 0.116 0.119 0.086 0.106
## Adjusted R2 0.116 0.112 0.114 0.078 0.097
## =======================================================================
## Note: *p<0.1; **p<0.05; ***p<0.0110 - Is there a variable that significantly changes the sample size of your regressions? Why
#Including age decreases the the sample size by about 200.
table(sharks$Age[is.na(as.numeric(sharks$Age))])## Warning in table(sharks$Age[is.na(as.numeric(sharks$Age))]): NAs introduced
## by coercion##
## 12 or 13 20s 30s 50s 60s 6½ a minor
## 198 1 3 3 1 2 1 1
## mid-30s teen Teen Teens
## 1 4 4 1#maybe we can start replacing teens with ~16, remove the "s" after some of these. Still, 198 missing all together.12 - Are serious injuries more or less likely given missingness for the variable?
lim <- is.na(as.numeric(sharks$Age))## Warning: NAs introduced by coerciont.test(sharks$severe_injury[lim], sharks$severe_injury[!lim])##
## Welch Two Sample t-test
##
## data: sharks$severe_injury[lim] and sharks$severe_injury[!lim]
## t = -5.0123, df = 305.38, p-value = 9.129e-07
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.247812 -0.108089
## sample estimates:
## mean of x mean of y
## 0.6318182 0.8097686t.test(sharks$westCoast[lim], sharks$westCoast[!lim])##
## Welch Two Sample t-test
##
## data: sharks$westCoast[lim] and sharks$westCoast[!lim]
## t = 3.2974, df = 318.61, p-value = 0.001086
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 0.04646003 0.18392090
## sample estimates:
## mean of x mean of y
## 0.3272727 0.2120823#So it appears that missing ages are less likely to be severly injured and more likely to be on the east coast. Perhaps this is spurious, but it might be a signal of something non-random.