Section 2.3 — Inventory of discrete distributions#
This notebook contains all the code examples from Section 2.3 Inventory of discrete distributions of the No Bullshit Guide to Statistics.
Notebook setup#
# load Python modules
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
# Figures setup
sns.set_theme(
context="paper",
style="whitegrid",
palette="colorblind",
rc={'figure.figsize': (6,4)},
)
%config InlineBackend.figure_format = 'retina'
# set random seed for repeatability
np.random.seed(42)
%pip install -q ministats
Note: you may need to restart the kernel to use updated packages.
from ministats import plot_pmf
from ministats import plot_cdf
Definitions#
Math prerequisites#
Combinatorics#
See SciPy docs:
https://docs.scipy.org/doc/scipy/reference/generated/scipy.special.factorial.html
https://docs.scipy.org/doc/scipy/reference/generated/scipy.special.perm.html
https://docs.scipy.org/doc/scipy/reference/generated/scipy.special.comb.html
Factorial#
from scipy.special import factorial
# ALT.
# from math import factorial
factorial(4)
24.0
factorial(1), factorial(2), factorial(3)
(1.0, 2.0, 6.0)
[factorial(k) for k in [5,6,7,8,9,10,11,12]]
[120.0, 720.0, 5040.0, 40320.0, 362880.0, 3628800.0, 39916800.0, 479001600.0]
factorial(15)
1307674368000.0
import numpy as np
np.log(factorial(15))/np.log(10)
12.116499611123398
Permutations#
from scipy.special import perm
perm(5,2)
20.0
perm(5,1), perm(5,2), perm(5,3), perm(5,4), perm(5,5)
(5.0, 20.0, 60.0, 120.0, 120.0)
Combinations#
from scipy.special import comb
comb(5,2)
10.0
Discrete distributions reference#
Discrete uniform#
# import the model family
from scipy.stats import randint
# choose parameters
alpha = 1 # start at
beta = 4 # stop at
# create the rv object
rvU = randint(alpha, beta+1)
# use one of the rv object's methods
The limits of the sample space of the random variable rvU
can be obtained by calling its .support()
method.
rvU.support()
(1, 4)
rvU.mean()
2.5
rvU.var()
1.25
rvU.std() # = np.sqrt(rvU.var())
1.118033988749895
Probability mass function#
for x in range(1,4+1):
print(f"f_U({x}) = ", rvU.pmf(x))
f_U(1) = 0.25
f_U(2) = 0.25
f_U(3) = 0.25
f_U(4) = 0.25
To create a stem-plot of the probability mass function \(f_U\), we can use the following three-step procedure:
Create a range of inputs
xs
for the plot.Compute the value of \(f_U =\)
rvU
for each of the inputs and store the results as list of valuesfUs
.Plot the values
fUs
by calling the functionplt.stem(xs,fUs)
.
import numpy as np
import matplotlib.pyplot as plt
xs = np.arange(0,8+1)
fUs = rvU.pmf(xs)
plt.stem(xs, fUs, basefmt=" ");
Cumulative distribution function#
for b in range(1,4+1):
print(f"F_U({b}) =", rvU.cdf(b))
F_U(1) = 0.25
F_U(2) = 0.5
F_U(3) = 0.75
F_U(4) = 1.0
import numpy as np
import seaborn as sns
xs = np.linspace(0,8,1000)
FUs = rvU.cdf(xs)
sns.lineplot(x=xs, y=FUs);
Let’s generate 10 random observations from random variable rvU
:
rvU.rvs(10)
array([3, 4, 1, 3, 3, 4, 1, 1, 3, 2])
Bernoulli#
from scipy.stats import bernoulli
rvB = bernoulli(p=0.3)
rvB.support()
(0, 1)
rvB.mean(), rvB.var()
(0.3, 0.21)
rvB.rvs(10)
array([0, 0, 1, 0, 1, 0, 1, 1, 0, 0])
Binomial#
We’ll use the name rvX
because rvB
was already used for the Bernoulli random variable above.
from scipy.stats import binom
n = 20
p = 0.14
rvX = binom(n,p)
rvX.support()
(0, 20)
rvX.mean(), rvX.var()
(2.8000000000000003, 2.4080000000000004)
Poisson#
from scipy.stats import poisson
lam = 10
rvP = poisson(lam)
rvP.pmf(8)
0.11259903214902009
rvP.cdf(8)
0.3328196787507191
## ALT. way to compute the value F_P(8) =
# sum([rvP.pmf(x) for x in range(0,8+1)])
Geometric#
from scipy.stats import geom
rvG = geom(p = 0.2)
rvG.support()
(1, inf)
rvG.mean(), rvG.var()
(5.0, 20.0)
Negative binomial#
from scipy.stats import nbinom
r = 10
p = 0.5
rvN = nbinom(r,p)
rvN.support()
(0, inf)
rvN.mean(), rvN.var()
(10.0, 20.0)
Hypergeometric (optional)#
from scipy.stats import hypergeom
a = 30 # number of success balls
b = 40 # number of failure balls
n = 20 # how many we're drawing
rvH = hypergeom(a+b, a, n)
rvH.support()
(0, 20)
rvH.mean(), rvH.var()
(8.571428571428571, 3.54924578527063)
Tomatoes salad probabilities#
Number of dogs seen by Amy#
a = 7 # number dogs
b = 20 - 7 # number of other animals
n = 12 # how many "patients" Amy will see today
rvD = hypergeom(a+b, a, n)
# Pr of exactly five dogs
rvD.pmf(5)
0.2860681114551084
Modelling real-world data using probability#
TODO: add simple inference and plots