Plot the Poisson CDF with the Standard Normal Distribution CDF

15 views (last 30 days)
Meng Fei Wong
Meng Fei Wong on 7 Jan 2022
Edited: Torsten on 8 Jan 2022
Hello all, I have a question on plotting the Poisson cdf together with the standard normal distribution cdf. Below is the task description:
Let X be a random variable with and be the normalized version of X.
The following figure shows the cdf of for some and the cdf Φ of the standard normal distribution.
I have tried my best to plot as same as the given figure but still I am not able to get it looks similiar as the given figure. Below is my result:
May I know what do I miss? Below is my code:
pe = makedist('Normal')
x1 = -3:.1:3;
p = cdf(pe,x1);
%plot(x1,p)
x3 = -3:3;
y3 = poisscdf(x3,1);
figure
stairs(x3,y3)
hold on
plot(x1,p)
xlabel('Observation')
ylabel('Cumulative Probability')

Answers (3)

VBBV
VBBV on 7 Jan 2022
Edited: VBBV on 7 Jan 2022
x3 = -3:0.1:3
Use the same intervals as x1
  1 Comment
John D'Errico
John D'Errico on 8 Jan 2022
Not the answer. A Poisson distribution will not have mass for negative values. So while making the increment for x3 does make sense to be consistent, it changes nothing.

Sign in to comment.


John D'Errico
John D'Errico on 8 Jan 2022
Edited: John D'Errico on 8 Jan 2022
Note that a Poisson random variable will ALWAYS be a non-negative number. It CANNOT have mass on the negative end of the real line. So the plot that you show, with a supposedly Poisson random variable that goes negative? It cannot exist. (At least not for a standard Poisson.)
Perhaps the intent is to have a Poisson random variable that has been transformed into the range of a normal distribution.
For example, Poisson distributions with large Poisson parameters will tend to look very normally distributed. (Not difficult to prove as I recall.)
fplot(@(x) poisscdf(x,50),[0,100])
So a Poisson CDF that does look quite normal. A quick glance at Wikipedia...
tells me that for a Poisson distribution with parameter lamnda, the mean will be lambda, as well as the variance. So we can simply do this:
lambda = 30;
fplot(@(x) poisscdf(x,lambda),[0,2*lambda])
hold on
fplot(@(x) normcdf(x,lambda,sqrt(lambda)))
grid on
hold off
s you should see, the two curves nearly overlay on top of each other.
In your figure, the Poisson was apparently transformed, via a transformation to look like a normal.
lambda = 30;
fplot(@(x) poisscdf(lambda + x*sqrt(lambda),lambda),[-3,3])
hold on
fplot(@(x) normcdf(x),[-3,3])
grid on
hold off
However, that is NOT a Poisson distibution because it is shown to have mass for negative x. It is derived from one.

Torsten
Torsten on 8 Jan 2022
Edited: Torsten on 8 Jan 2022
I suspect the graphics shows that the arithmetic mean of normalized independent Poisson random variables converges in distribution to the standard normal distribution.
You can see it from the following plot:
function main
lambda = 3;
m = 2000; % number of RVs
n = 3000; % number of samples
a = poissrnd(lambda,m,n);
for i=1:n
b(i) = (sum(a(:,i))/m-lambda)/sqrt(lambda/m); % normalized arithmetic mean of Poisson RVs
end
cdfplot(b); % plot empirical cdf of b
hold on
x = linspace(min(b),max(b));
plot(x,normcfd(x,0,1)) % compare with standard normal distribution
end

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!