You are on page 1of 4

Roberto Valdez

Uni: rjv2119
STAT W4290

Homework 1

Problem 1)

Using R we plot the required graph:

We can see that GM returns and Ford returns in the large majority are concentrated in the center and
show a somewhat positive correlation. The outliers do appear together (high values for GM imply high
values for Ford and otherwise).

Problem 5,6,7)

We simulate the returns and we include stopping points when the daily price goes over 1,100,000 or
under 950,000. Then we obtain:

Pr(𝐿𝑜𝑠𝑠) = 0.58844

mean(𝑃𝑟𝑜𝑓𝑖𝑡) = 9,922.63

mean(𝐷𝑎𝑖𝑙𝑦_𝑅𝑒𝑡𝑢𝑟𝑛) = −0.01783836

We include an R file with the code, modified from the one provided in the textbook that we used for
obtaining this results.
Exercise 1 a)

We are going to express 𝛷(𝑥) and 𝜑(𝑥) as the C.D.F. and P.D.F. respectively, of a (0,1) Gaussian
distribution. We do:
ln 𝑦 − 𝜇
Pr(𝑌 ≤ 𝑦) = Pr(𝑒 𝑋 ≤ 𝑦) = Pr(𝑋 ≤ ln 𝑦) = 𝛷 ( )
𝜎

Now we have that:

ln 𝑦−𝜇 2
−( )
𝑑 𝑑 ln 𝑦 − 𝜇 1 ln 𝑦 − 𝜇 1 1 𝜎
𝑓𝑌 (𝑦) = 𝐹𝑌 (𝑦) = 𝛷( )= 𝜑( )= ( 𝑒 2 )
𝑑𝑦 𝑑𝑦 𝜎 𝜎𝑦 𝜎 𝜎𝑦 √2𝜋

1 −(ln 𝑦−𝜇)2
= 𝑒 2𝜎 2
√2𝜋𝜎𝑦

Exercise 1 b)

𝔼[𝑌] = 𝔼[𝑒 𝑋 ]

And this equals the moment generating function of 𝑋 with 𝑡 = 1. Therefore:

1 2
𝔼[𝑌] = 𝔼[𝑒 𝑋 ] = 𝑀. 𝐺. 𝐹(1) = 𝑒 𝜇+2𝜎

Now to get the variance we need the second moment and therefore:

𝔼[𝑌 2 ] = 𝔼[𝑒 2𝑋 ]

And this equals the moment generating function of 𝑋 with 𝑡 = 2. Therefore:


2
𝔼[𝑌 2 ] = 𝔼[𝑒 2𝑋 ] = 𝑀. 𝐺. 𝐹(1) = 𝑒 2𝜇+2𝜎

We compute the variance as:


2 2 2 2
𝑉𝑎𝑟(𝑌) = 𝔼[𝑌 2 ] − 𝔼2 [𝑌] = 𝑒 2𝜇+2𝜎 − 𝑒 2𝜇+𝜎 = 𝑒 2𝜇+𝜎 (𝑒 𝜎 − 1)
Exercise 2 a)

We know that if
𝐼(𝑋𝑖 ≤ 𝑥)~𝐵𝑒𝑟𝑛𝑜𝑢𝑙𝑙𝑖(𝑝 = 𝐹(𝑥))

Then:
𝑛

𝑌 = ∑ 𝐼(𝑋𝑖 ≤ 𝑥) ~𝐵𝑖𝑛𝑜𝑚𝑖𝑎𝑙(𝑛, 𝑝 = 𝐹(𝑥))


𝑖=1
𝔼[𝑌] = 𝑛𝐹(𝑥)

Var(𝑌) = 𝑛𝐹(𝑥)(1 − 𝐹(𝑥))

Then:
𝑛
1 𝑌 1
𝔼[𝐹𝑛 (𝑥)] = 𝔼 [ ∑ 𝐼(𝑋𝑖 ≤ 𝑥)] = 𝔼 [ ] = 𝔼[𝑌] = 𝐹(𝑥)
𝑛 𝑛 𝑛
𝑖=1

And with this we have shown that 𝐹𝑛 (𝑥) is an unbiased estimator for 𝐹(𝑥).

Exercise 2 b)

With what was stated above we can do:


𝑛
1 𝑌 1 𝐹(𝑥)(1 − 𝐹(𝑥))
Var(𝐹𝑛 (𝑥)) = Var ( ∑ 𝐼(𝑋𝑖 ≤ 𝑥)) = Var ( ) = 2 Var(𝑌) =
𝑛 𝑛 𝑛 𝑛
𝑖=1

Exercise 2 c)

Knowing that:
𝐹(𝑥)(1 − 𝐹(𝑥))
𝐹𝑛 (𝑥)~𝐹 ∗ (𝐹(𝑥), )
𝑛

Stating that 𝐹 ∗ is a distribution. Using the central limit theorem we know that for a large n:

√𝑛(𝐹𝑛 (𝑥) − 𝐹(𝑥))


~𝑁(0,1)
√𝐹(𝑥)(1 − 𝐹(𝑥))
Appendix (R code):

for (i in 1:niter)
{
r = rnorm(100,mean=.05/253,
sd=.23/sqrt(253)) # generate random numbers
logPrice = log(1e6) + cumsum(r)
#We generate a sequence that stops if a min or a max is obtained
pmax = ifelse(length(which(logPrice >= log(1100000))) > 0,
min(which(logPrice >= log(1100000))), 100)
pmin = ifelse(length(which(logPrice < log(950000))) > 0,
min(which(logPrice < log(950000))), 100)
daystrade = min(pmax,pmin)
logPrice2 =logPrice[1:daystrade]
proof[i] = min(logPrice2)
minlogP = min(logPrice2) # minimum price over next 100 days
below[i] = as.numeric(minlogP < log(950000))
if(daystrade == 100){
profit[i] = max(min(exp(logPrice2[100]) - 1.e6,100000),-50000)
}
else{
if(pmax < pmin){
profit[i] = 100000
}
else{
profit[i] = -50000
}
}
sim_return[i] = profit[i] / 50000.
sim_return[i] = sim_return[i] / daystrade # convert into daily returns
}

mean(below)
mean(profit)
mean(sim_return)

You might also like