In probability theory, the central limit theorem (CLT) establishes that, when independent random variables are added together, their properly normalized sum tends toward a normal distribution (informally a “bell curve”) even if the original variables themselves are not normally distributed. To understand this theorem, suppose you generate 100 uniform random numbers and sum them to get a single number. Then you repeat this procedure 1000 times to get 1000 of these sums of 100 uniform random numbers. The CLT theorem implies that if you plot a histogram of the values of these 1000 sums, then the resulting distribution looks very much like the Gaussian bell-shaped function. The larger the number of these sums (for example, 100000 instead of 1000 sums), the more the resulting distribution will look like a Gaussian. Here we want to see this theorem in action.
Consider a random walker, who takes a random step of a uniformly-distributed random-size between $[0,1]$, in positive or negative directions on a single staright line. The random walker can repeat these steps for nstep times, starting from an arbitrary initial starting point.
Write a function with the interface doRandomWalk(nstep,startPosition), that takes the number of steps nstep for a random walk and the startPosition of the random walk on a straight line, and returns the location of the final step of the random walker.
Now, write another function with the interface simulateRandomWalk(nsim,nstep,startPosition) that simulates nsim number of random-walks, each of which contains nstep steps and starts at startPosition. Then, this function calls doRandomWalk() repeatedly for nsim times and finally returns a vector of size nsim containing final locations of all of the nsim simulated random-walks.
Now write a script that plots the output of simulateRandomWalk() for
The resulting plot should look like the following,
How do you interpret this result? How can uniformly-distributed random final steps end up having a Gaussian bell-shape distribution.
The reason for the resulting Gaussian-looking distribution function is that the positions of the final steps are the results of additions of a fixed number of identically-distributed random-variables (i.e., the 10 random steps in each random-walk). Therefore, by the CLT, the resulting distribution of the final steps should resemble the Gaussian distribution.
def doRandomWalk(nstep,startPosition):
"""
Returns the final location of nstep random-walk steps on a straight line.
"""
import numpy as np
import numpy.random as rnd
lastStepPosition = startPosition + np.sum( ( rnd.random(nstep) ) * 2 - 1 )
return lastStepPosition
def simulateRandomWalk(nsim=10000,nstep=100,startPosition=0):
import numpy as np
LastStepVec = np.zeros(nsim)
for i in range(nsim):
LastStepVec[i] = doRandomWalk(nstep,startPosition)
return LastStepVec
import matplotlib.pyplot as plt
fig = plt.figure( figsize=(16, 9) \
, dpi= 300 \
, facecolor='w' \
, edgecolor='w' \
) # create figure object
ax = fig.add_subplot(1,1,1) # Get the axes instance
nsim = 10000
nstep = 10
startPosition = -10
ax.hist ( simulateRandomWalk(nsim=nsim,nstep=nstep,startPosition=startPosition) \
, alpha=0.5 \
)
ax.set_xlabel('Last-Step Position')
ax.set_ylabel('Count')
ax.set_title('Histogram of last steps in {} Random Walk Simulations of {} Steps Starting at {}'.format(nsim,nstep,startPosition))
fig.savefig('randomWalk1D.png', dpi=100) # save the figure to an external file
plt.show() # display the figure
