knitr::opts_chunk$set(echo = TRUE,
fig.pos='H')
directory = "C:/Users/Michael/Dropbox/priv/CUNY/MSDS/201909-Fall/DATA605_Larry/20190929_Week05/"
knitr::opts_knit$set(root.dir = directory)
### Make the output wide enough
options(scipen = 999, digits=12, width=150)
### no other libraries required this weekp.59: Definition 2.1 Let \(X\) be a continuous real-valued random variable.
A density function for \(X\) is a real-valued function \(f\) which satisfies
\[P(a \le X \le b) = \int_a^b f(x) dx\] for all \(a,b\in R\) .
The uniform PDF on the interval [0,1] is defined as \[f_X(x)= \left\{ \begin{matrix} 0, & if\quad x\le0 \\ 1, & \quad if\quad 0 \le x \le1 \\ 0, & if\quad x \ge1 \end{matrix} \right. \] Thus, the density is always non-negative, as required.
For the uniform density function on the interval [0,1], \(f(x) = 1\), so \[P(a \le X \le b) = \int_a^b 1\cdot dx = b-a\] for \(0 \le a \le b \le 1\) .
P.60: The probability that the outcome of the experiment falls in an interval \([a, b]\) is given by
\[P([a, b]) = \int_a^b f(x)dx\] ,
that is, by the area under the graph of the density function in the interval \([a, b]\).
For the interval [0,1], the probability is \[P(0 \le X \le 1) = \int_0^1 1\cdot dx = 1-0 = 1\] thus the total probability equals 1, as required.
P.61: Definition 2.2 Let X be a continuous real-valued random variable.
Then the cumulative distribution function (CDF) of X is defined by the equation \(F_X(x) = P(X \le x)\) .
It is clear that X always takes on a value between 0 and 1, so the uniform cumulative distribution function of X is given by
\[F_X(x)= \left\{ \begin{matrix} 0, & if\quad x\le0 \\ x, & \quad if\quad 0 \le x \le1 \\ 1, & if\quad x \ge1 \end{matrix} \right. \]
The distribution is always non-negative, as required.
Since \(B\) and \(C\) are independent and identially distributed to \(X\), the above applies to each of \(B,C\).
Note that the point (B,C) is then chosen at random in the unit square. Find the probability that:
The sum of two uniform variables corresponds to example 2.14 on pp. 63-64.
Let \(Z = B+C\). Then the PDF for \(Z\) has positive density on the interval [0,2] and is defined as \[f_Z(z)= \left\{ \begin{matrix} 0, & if\quad z\le0 \\ z, & \quad if\quad 0 \le z \le 1 \\ 2-z, & \quad if\quad 1 \le z \le 2 \\ 0, & if\quad z \ge 2 \end{matrix} \right. \]
Density of the sum z is the large triangle from (0,0) to (1,1) to (2,0):
The total area under this (large) triangle is \(\frac{width * height}{2} = 2 \cdot 1 \cdot \frac{1}{2}=1\) .
The CDF for \(Z\) has positive density on the interval [0,2] and is defined as \[f_Z(z)= \left\{ \begin{matrix} 0, & if\quad z\le0 \\ \frac{z^2}{2}, & \quad if\quad 0 \le z \le 1 \\ 1-\frac{(2-z)^2}{2}, & \quad if\quad 1 \le z \le 2 \\ 1, & if\quad z \ge 2 \end{matrix} \right. \]
The region in which density \(=z < \frac{1}{2}\) is the smaller triangle from (0,0) to (\(\frac{1}{2}\),0) to (\(\frac{1}{2}\),\(\frac{1}{2}\)), shaded in yellow.
The area of this yellow triangle is \(\frac{width * height}{2} = \frac{1}{2} \cdot \frac{1}{2} \cdot \frac{1}{2}=1(\frac{1}{2})^3 = 0.125 = \frac{1}{8}\) , which can be computed from the CDF for \(z=\frac{1}{2}\) . Therefore, the theoretical probability is \(\frac{1}{8}=0.125\) .
n=1000000
B=runif(n,0,1)
C=runif(n,0,1)
Z=B+C
prob1 = sum(Z<=1/2)/n
# Probability that Z=B+C < 1/2 :
prob1## [1] 0.124744
The simulated probability that \(Z=B+C < \frac{1}{2}\) is 0.124744 .
This has to be separated into two cases, the results of which will be summed for the final answer:
if \(B \le \frac{1}{2}\) then \(C\) can take on any value \(\in [0,1]\) and \(B \cdot C \le \frac{1}{2}\) .
On a grid plotting B vs. C, this will give a rectangle of width \(\frac{1}{2}\) and height 1 for an area of \(\frac{1}{2}\) .
if \(B > \frac{1}{2}\) then we must have \(C < \frac{0.5}{x}\) in order for \(B \cdot C \le \frac{1}{2}\) .
On a grid plotting B vs. C, for \(B \in [\frac{1}{2},1]\) this will give the following curve for C:
\[Z = \int\limits_{b=\frac{1}{2}}^{b=1} \left[ \frac{0.5}{c} \right] dc = \frac{1}{2} \log [c] \left| \begin{matrix} b=1 \\ & \\ b=\frac{1}{2}\end{matrix} \right. = \frac{1}{2} \left[log(1) - log\left(\frac{1}{2}\right)\right] = \frac{1}{2} \left[0 - log\left(\frac{1}{2}\right)\right] = \frac{1}{2} \left[log\left(2)\right)\right] = log(sqrt(2))\]
So, the theoretical answer is \(Pr \left( Z=B \cdot C \le \frac{1}{2}\right) = \frac{1}{2} + \ln(\sqrt{2}) \approx 0.846574\)
n=1000000
B=runif(n,0,1)
C=runif(n,0,1)
Z=B*C
prob2 = sum(Z<=1/2)/n
# Probability that Z=B*C < 1/2 :
prob2## [1] 0.846637
The simulated probability that \(Z=B \cdot C < \frac{1}{2}\) is 0.846637 .
This corresponds to Example 2.16 on pp.65-66, and figure 2.19 on p.67:
Let \(Z = |B-C|\). Then the PDF for \(Z\) has positive density on the interval [0,1] and is defined as \[f_Z(z)= \left\{ \begin{matrix} 0, & if\quad z\le0 \\ 2(1-z), & \quad if\quad 0 \le z \le 1 \\ 0, & if\quad z > 1 \end{matrix} \right. \]
Then the CDF for \(Z\) is defined as \[F_Z(z)= \left\{ \begin{matrix} 0, & if\quad z\le0 \\ 1-(1-z)^2, & \quad if\quad 0 \le z \le 1 \\ 1, & if\quad z > 1 \end{matrix} \right. \]
For \(z=\frac{1}{2}\) , the theoretical probability that \(Pr \left( |B - C| < z=\frac{1}{2} \right)\) is \[ \begin{aligned} Pr \left(z=|B-C| \le \frac{1}{2} \right) = 1-\left(1-\frac{1}{2}\right)^2 = 1-\left(\frac{1}{2}\right)^2 = 1-\frac{1}{4} = \frac{3}{4} = 0.75 \end{aligned} \].
Grid of B vs C where |B-C| is less than z
Removing the triangles of size \(\frac{1}{8}\) from the square above yields the same result: \(1 - \frac{1}{8} - \frac{1}{8} = \frac{6}{8} = \frac{3}{4} = 0.75\) .
n=1000000
B=runif(n,0,1)
C=runif(n,0,1)
Z=abs(B-C)
prob3 = sum(Z<=1/2)/n
# Probability that Z=|B-C| < 1/2 :
prob3## [1] 0.750082
The simulated probability that \(Z= |B-C| < \frac{1}{2}\) is 0.750082 .
\[ \begin{aligned} Pr \left( Z=max(B,C) < \frac{1}{2} \right) &= Pr \left( B< \frac{1}{2} \wedge C< \frac{1}{2} \right) \\ &= Pr \left( B< \frac{1}{2}\right) \cdot Pr \left( C< \frac{1}{2} \right) \\ &= \frac{1}{2} \cdot \frac{1}{2} \\ &= \frac{1}{4} \\ &= 0.25 \end{aligned} \]
where the second equality comes from independence of \(B\) and \(C\) .
The theoretical probability that \(Z=max(B,C) < \frac{1}{2}\) is \(\frac{1}{2} \cdot \frac{1}{2} = \frac{1}{4} = 0.25\) .
n=1000000
B=runif(n,0,1)
C=runif(n,0,1)
Z1=cbind(B,C)
Z=apply(Z1,1,max)
prob4 = sum(Z<=1/2)/n
# Probability that Z=|B-C| < 1/2 :
prob4## [1] 0.249115
The simulated probability that \(Z=max(B,C) < \frac{1}{2}\) is 0.249115 .
\[ \begin{aligned} Pr \left( Z=min(B,C) < \frac{1}{2} \right) &= Pr \left( \left( B < \frac{1}{2}\right) \vee \left( C< \frac{1}{2} \right) \right) \\ &= 1 - Pr \left( \left( B > \frac{1}{2} \right) \wedge \left(C > \frac{1}{2} \right) \right) \\ &= 1 - Pr \left( B > \frac{1}{2}\right) \cdot Pr \left( C > \frac{1}{2} \right) \\ &= 1 - \frac{1}{2} \cdot \frac{1}{2} \\ &= 1 - \frac{1}{4} \\ &= 0.75 \end{aligned} \]
where the second equality stems from De Morgan’s Law,
https://en.wikipedia.org/wiki/De_Morgan%27s_laws
and the third equality steps from independence of \(B\) and \(C\) .
The theoretical probability that \(Z=max(B,C) < \frac{1}{2}\) is \(1 - \frac{1}{2} \cdot \frac{1}{2} = 1 - \frac{1}{4} = \frac{3}{4} = 0.75\) .
n=1000000
B=runif(n,0,1)
C=runif(n,0,1)
Z1=cbind(B,C)
Z=apply(Z1,1,min)
prob5 = sum(Z<=1/2)/n
# Probability that Z=|B-C| < 1/2 :
prob5## [1] 0.749826
The simulated probability that \(Z=min(B,C) < \frac{1}{2}\) is 0.749826 .