Random
Numbers and Simulations

We’ll use Matlab to generate random
numbers so that we can discuss its statistical capabilities
and explore some
simulations, too. We’ll start with a definition of some
statistical quantities.
In these sections we consider a set of data which we’ll refer to as x_{i} where
i
= 1 to N.
We’ll not make any assumptions about the data in terms of its source or
its form.

Averages
The mean is given by adding up all the data and dividing by the number
of objects in the set. This is written as:
There’s a Matlab command
which does exactly this: ‘mean(x)’.
There’s another related
command, which is ‘median(x)’
and it’s
the value in the ‘middle’ of a vector when the data has been sorted. If there’s
an odd number of pieces of data this is well defined, however if the
number is
even then the mean of the data at each end of the central interval is
given,
for example the median of [1 2 3 4] is (2+3)/2 = 2.5 whereas the median
of [1 3
5] is 3. Note that the median can equal the mean but it is not
necessarily so.
This short code
x = [4 0 5 3 2 1 3 5 9 3 7 5 6 8 2 0];
mn = mean(x)
md = median(x)
results in
mn = 3.9375
md = 3.5000
Other Statistical Measures
In the previous section
we have discussed averages of a set
of data but there are more measures available, for instance these two
distributions have the same mean but are obviously different in form:
The first curve is
narrower than the second and this is
quantified using the variance, which is given by
We can interpret this value as the ‘mean’
of the sum of the
squares of the distance from the mean. There’s the Matlab command ‘var(x)’ to
calculate this number. Another related measure is the standard deviation,
which
is the square root of the variance, ‘std(x)’.
This also is a measure of the
width of the distribution and has the advantage that it has the same
units as
the data.
There are other measures, some of which are called
‘higherorder moments’
(mean is
first and variance
is second): the skewness
is the
third, and the kurtosis
is the fourth.
Random Numbers and
Distributions
In order to generate
random numbers we can use various
commands available
in Matlab. We won’t worry about how this is done or the
technical aspects of the seeding. We’re going to start with the simple
command
‘rand’ which
returns a random number between zero and one.
If we rolled a die a
large number of times, and we used the
above formulas to calculate the mean and variance, we’d expect to
obtain a mean
= 3.5 and a variance = 2.916. Let’s try it with Matlab.
This is a simple dice program:
function n =
roll_d()
n =
ceil(rand*6 + eps);
We can use it like this:
%
Roll the die and keep the values in d
for i = 1 :
1000
d(i)
= roll_d;
end
%
Find how many appearences of each possible value
for i = 1 :
6
v(i)
= length(find(d==i));
end
%
Display results
disp(['Mean
= '
num2str(mean(d))])
disp(['Variance
= '
num2str(var(d))])
bar(v)
The results are:
Mean = 3.508
Variance = 2.7827
Obviously, every time
that we run the program we’ll get
different results, since we are dealing with random numbers, but the
results
are similar to what we expected.
We can use this process
for other ‘chance games’,
for
example tossing a coin:
function r =
toss()
p = ['heads'; 'tails'];
i =
ceil(rand*2 + eps);
r =
p(i,:);
This can then be called just using 'toss'
which generates
either 'heads' or 'tails'.
Normal Distribution
A normal distribution has
two parameters associated with it:
the mean and the variance. The command ‘randn’ generates
random
numbers which have a mean
of zero and a variance
of unity.
By typing
x = randn(1, 1000);
hist(x,12)
we set up an array and
produce the plot
These
lines on the command window
mn = mean(x)
v =
var(x)
produce, in our case
mn = 0.0016
v =
0.9908
which
are values very close to what was expected.
There
are many problems which require simulation, largely due to
that it's not viable to actually run the real or physical
tests.
These cases cover things from experiments on aircrafts or cars
(which are too dangerous or expensive) to population studies and
economic strategies.
From
'Random Numbers' to
home
From
'Random Numbers' to 'Probability and Statistics'
