1
$\begingroup$

Forgive me I am a newbie of random variables. I saw a lot of course which introduce the Discrete Random Variables which always be illustrated with a histogram or bar chart . Like below

enter image description here

But in my understanding .I think it should be represented by a graph like below, all the probability value for the lowercase x {2,3,4,5} is point. not an area. (forgive me bad drawing).

My question is why it should be illustrated by a histogram or bar chart instead of point? Please correct me if there is something wrong . Thanks.

enter image description here

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ See Glen_b's answer to this question $\endgroup$
    – user20160
    Commented Jan 21, 2019 at 18:18

2 Answers 2

Reset to default
1
$\begingroup$

It is just a standard way to plot the distributions.

Theoretically, a point is more adequate to symbolize a single numerical value, e.g. P(x=2) = 2/14. Nonetheless, and despite not having an intrinsic value within themselves, it is visually easier and quicker to represent the values with bars.

Improve this answer
$\endgroup$
3
  • 1
    $\begingroup$ As the answer by BruceET shows, there is a point to using histograms (which represent probability by area): these have a direct connection with probability density functions, which also represent probability by area. Their use for discrete variables is not strictly necessary, but is an excellent conceptual and pedagogical tool for generalizing results and intuition about discrete variables to continuous variables. $\endgroup$
    – whuber ♦
    Commented Jan 21, 2019 at 18:25
  • $\begingroup$ @whuber Thanks your comments, for better understanding. It will be appreciated if you can kind help to explain more about probability density functions. Thanks in advance. +1 $\endgroup$
    – Joe.wang
    Commented Jan 22, 2019 at 15:13
  • $\begingroup$ I think understand direct connection with probability density functions is the key. thanks. $\endgroup$
    – Joe.wang
    Commented Jan 22, 2019 at 15:16
1
$\begingroup$

Showing the normal approximation to $\mathsf{Binom}(16, .5).$ 

x = 0:16;  pdf = dbinom(0:16, 16, .5)
hdr = "PDF of BINOM(16, .5) with Normal Approx"
plot(x, pdf, type="h", lwd=2, col="blue", main=hdr)
curve(dnorm(x, 8, 2), lwd=2, col="red", add=T)
abline(h=0, col="green2")

enter image description here

If $\mathsf{Binom}(16, .5)$ is approximated by data then a histogram on a density scale may be a more appropriate graphical display. A sample of size 10,000 approximates the binomial model fairly well, but not perfectly. The histogram shows the simulated data data. Centers of open circles show exact binomial probabilities.

set.seed(2019)
X = rbinom(10^4, 16, .5)
table(X)
X
   1    2    3    4    5    6    7    8    9   10   11   12   13   14   15 
   5   29   92  300  675 1265 1743 1936 1757 1195  630  272   75   23    3 

hdr2 = "Histogram of large normal sample"
hist(X, prob=T, br = (-1:16)+.5, col="skyblue2")
 points(0:16, pdf, col="red")

enter image description here

Note: (a) Graphs made using R statistical software. (b) As illustrated in @Glen_b's link, a plot of the CDF of a discrete distribution may sometimes be preferable.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.