DATA 352/551: Homework 3

Question 1

Suppose we roll a pair of fair dice. A gambler bets that, in 10 rolls, a pair of six would show up. What is his chance of winning? We wrote down the four steps to carry out this simulation in Lecture 1 Example 3. Write code for this simulation and estimate his chance of winning.

Show solution
# 1. ONE simulation function
sim_dice_pair <- function(){
  roll1 <- sample(x=c(1:6),size=10,replace=T)
  roll2 <- sample(x=c(1:6),size=10,replace=T)
  
  ## save win/lose in each simulation
  if(max(roll1+roll2)==12){
    return('win')   # win if pair of six showed up
  }else{
    return('lose')
  }
}

# 2. Test the function
sim_dice_pair()
[1] "lose"
# 3. Initialize global simulation parameters
NSIM <- 1000
result <- rep(NA,NSIM)

# 4. Repeat the simulation NSIM times

#### Method 1: 'for' loop
for(i in 1:NSIM){
  result[i] <- sim_dice_pair()
}

#### Method 2: Base R replicate()
result <- replicate(NSIM, sim_dice_pair())

# 5. Summarize results
table(result)/NSIM
result
 lose   win 
0.776 0.224 
## to compare: exact probability
1-(35/36)^10
[1] 0.2455066

Question 2

You play a game with me as follows. A fair coin is tossed. If a head shows up, I give you a dollar. If a tail shows up, you give me a dollar. Suppose you keep playing this game until you win \(\$5\). Let \(Y\) be the length of the game you have to play. What is the distribution of \(Y\)? We wrote down the four steps to carry out this simulation in Lecture 1 Example 4b. Write code for this simulation and produce a histogram to estimate the distribution of \(Y\). Caution: as it is possible that you “never” reach \(\$5\), it is important to set an upperbound to the length of the game (i.e., if the length of the game has reached 10000, say, you quit the while loop regardless of your winning.)

The following code visualizes one round of the game, for demonstration purposes.

## global constant and initialization
target <- 5  # target winning
num.game.max <- 10000 # maximum game length for the while loop 
current  <- 0 # current net winning
num.game <- 0 # current game length
current.trace <-  current # winning trajectory for plotting

while((current<target)&(num.game<num.game.max)){
  num.game <- num.game+1
  current <- current + sample(c(-1,1),size=1,replace=T)
  current.trace <- c(current.trace,current)
}

# visualize result
if(num.game>=num.game.max){
  mytitle='did not reach target in 10000 runs'
}else{
  mytitle=paste('target reached in',num.game,'runs')
}
plot(current.trace,type='b',main=mytitle,
     xlab='number of runs',
     ylab='winnings')
abline(h=target,lty='dotted',col='red')

Show solution
# 1. ONE simulation function
sim_game_length <- function(target=5, num.game.max=10000){
  current <- 0 # current net winning
  num.game <- 0 # current game length

  while((current<target)&(num.game<num.game.max)){
    num.game <- num.game+1
    current <- current + sample(c(-1,1),size=1,replace=T)
  }
  return(num.game)
}

# 2. Test the function
sim_game_length()
[1] 343
# 3. Initialize global simulation parameters
NSIM <- 1000
num.game.max <- 10000 # maximum game length for the while loop
result <- rep(NA,NSIM)

# 4. Repeat the simulation NSIM times

for(i in 1:NSIM){
  result[i] <- sim_game_length(num.game.max=num.game.max)
  
  if(i %% 1000 == 0){
    print(paste(Sys.time(),'...running simulation',i))
  }
}
[1] "2026-09-23 15:52:12.858851 ...running simulation 1000"
# 5. summarize results
summary(result)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
    5.0    21.0    59.0   679.4   225.0 10000.0 
hist(result,nclass=50)

# we can take log transformation when histogram is severely right skewed
hist(log(result,base=10),nclass = 50)

mean(result==num.game.max) #prob. that game length >= num.game.max
[1] 0.03

Question 3

Write code to carry out the simulation in Homework 1, question 2.

Suppose you roll three fair dice and calculate the sum of the three rolls. Is the sum more likely to be 9 or 10? You can simulate the result of three dice rolls on random.org/dice for a few times to get a sense. Use the four steps discussed in the lecture to describe how to estimate the probability that the sum is 9 and the probability that the sum is 10.

Show solution
# 1. ONE simulation function
sim_three_dice <- function(){
  dices <- sample(c(1:6),size=3,replace=T) #sample 3 dices
  return(sum(dices)) #store the sum of dices
}

# 2. Test the function
sim_three_dice()
[1] 13
# 3. Initialize global simulation parameters
NSIM <- 1000
result <- rep(NA,NSIM)

# 4. Repeat the simulation NSIM times
result <- replicate(NSIM, sim_three_dice())

# 5. Summarize results
plot(as.factor(result),main='sum of three dice rolls')

Question 4

Write code to carry out the simulation in Homework 1, question 4.

The psychologist Tversky and his colleagues say that about four out of five people will answer (a) to the following question:

A certain town is served by two hospitals. In the larger hospital about 45 babies are born each day, and in the smaller hospital 15 babies are born each day. Although the overall proportion of boys is about 50 percent, the actual proportion at either hospital may be more or less than 50 percent on any day. At the end of a year, which hospital will have the greater number of days on which more than 60 percent of the babies born were boys?

  1. the large hospital
  2. the small hospital
  3. neither: the number of days will be about the same.

Assume that the probability that a baby is a boy is 0.5. Decide, by simulation, what the right answer is and suggest why so many people go wrong.

Show solution
# 1. ONE simulation function
sim_hospital_year <- function(births_per_day){
  year <- rep(NA,365) #to store percentage of boys each day
  for(d in 1:365){
    outcome <- sample(c(0,1),size=births_per_day,replace=T)
    year[d] <- mean(outcome)
  }
  return(sum(year>=0.6)) #number of days with >=60% boys
}

# 2. Test the function
sim_hospital_year(births_per_day=45)
[1] 47
sim_hospital_year(births_per_day=15)
[1] 112
# 3. Initialize global simulation parameters
NSIM <- 10000  #number of years to simulate
result.large <- rep(NA,NSIM) 
result.small <- rep(NA,NSIM)

# 4. Repeat the simulation NSIM times

result.large <- replicate(NSIM, sim_hospital_year(births_per_day=45))
result.small <- replicate(NSIM, sim_hospital_year(births_per_day=15))

# 5. summarize results
summary(result.large)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
  20.00   38.00   42.00   42.48   47.00   67.00 
summary(result.small)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   79.0   105.0   111.0   110.8   117.0   142.0 
# visualize results
par(mfrow=c(1,2))
hist(result.large, xlab='Num days with >=60% boys', 
     col='salmon',
     main='large hospital')
hist(result.small, xlab='Num days with >=60% boys', 
     col='lightblue',
     main='small hospital')

Question 5

Suppose that when cars drive past Drew toward downtown from 7am to 8am on a typical Sunday morning, \(41\%\) will turn onto Kings Rd., \(38\%\) will continue straight to Main St., and \(21\%\) will turn onto Park Ave. Assume that the number of cars driving past Drew toward downtown during that hour follows a Poisson distribution with parameter \(\lambda=6\). Let \(X\) be the number of cars that will drive on Main St. past the fork. Use simulation to answer the following questions.

  1. What is the probability that 5 or more cars will drive on Main street past the fork from 7am to 8am on a typical Sunday morning (i.e., \(P(X\ge5)\))?
  2. Estimate the distribution of \(X\) by plotting a histogram.
Show solution
# 1. ONE simulation function
sim_cars <- function(lambda=6, prob=0.38){
  numcar <- rpois(1,lambda=lambda)
  return(rbinom(1,size=numcar,prob=prob))
}

# 2. Test the function
sim_cars()
[1] 4
# 3. Initialize global simulation parameters
NSIM <- 1000
result <- rep(NA,NSIM)

# 4. Repeat the simulation NSIM times
result <- replicate(NSIM, sim_cars())

# 5. summarize results
mean(result>=5)
[1] 0.093
plot(as.factor(result),main='Number of cars that go to Main Street',col='salmon')