Scribd
Upload
20 views4 pages

Class Notes v1

Uploaded by

Parvat Chavhan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
20 views4 pages

Class Notes v1

Uploaded by

Parvat Chavhan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Probability, Distributions and Descriptive Statistics

Experiment
One-off or repeated process, for which possible outcomes are known, but actual outcome of each process is
known with some probability. The set of all possible outcomes is known as the sample space. The set of all
probabilities is known as the probability space; each element in the probability space lies between 0 and 1, and
the sum of all elements is one.

Random Variable
A function that maps a real number from the sample space to the probability space.

Probability Density Function


The set of ordered pairs (x, f(x)) is a probability density function (pdf) if:
1. 𝑓(𝑥) ≥ 0, ∀ 𝑥 ∈ ℝ
"
2. ∑! 𝑓(𝑥) = 1 if x is discrete, or, ∫#" 𝑓(𝑥) = 1 if x is continuous
$
3. 𝑃(𝑋 = 𝑥) = 𝑓(𝑥) if x is discrete, or, 𝑃(𝑎 < 𝑋 < 𝑏) = ∫% 𝑓(𝑥). 𝑑𝑥 if x is continuous

The pdf of a variable x, is commonly referred to as the distribution of x.

Cumulative Density Function


The cumulative density function (cdf), F(x), for a random variable X, with a probability density function f(x), is
!
defined as 𝐹(𝑥) = 𝑃(𝑋 ≤ 𝑥) = ∑&'! 𝑓(𝑡) if x is discrete, or, 𝐹(𝑥) = 𝑃(𝑋 ≤ 𝑥) = ∫#" 𝑓(𝑥). 𝑑𝑥 if x is continuous.

Expected Value of a Random Variable


Expected value is a generalised form of the weighted arithmetic average. Essentially, it tells one what the centre of
"
a distribution is. It is defined as: 𝐸(𝑋) = 𝜇 = ∑! 𝑥. 𝑃(𝑥) if x is discrete; and, 𝐸(𝑋) = 𝜇 = ∫#" 𝑥. 𝑃(𝑥). 𝑑𝑥 if x is
continuous.

Expected Value of a Linear Combination of Random Variables


𝐸(𝑎𝑋 + 𝑏𝑌 + ⋯ ) = 𝑎. 𝐸(𝑋) + 𝑏. 𝐸(𝑌) + ⋯

Variance
A measure of spread of variable x around its mean: 𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 ( ) − (𝐸(𝑋))(

Standard Deviation
𝜎(𝑋) = C𝑉𝑎𝑟(𝑋)

Range, Minimum, Maximum


Range defines the maximum and minimum points of an ordered set of all observations of the variable x.

Median, Percentiles, Deciles, Quartiles


For a variable x, the mid-point of its ordered set is called the median.

Essentially, if all the observations for the variable x are ordered, then the percentile level n, gives the percentage of
all observations that exist below that level. In other words, 90th percentile is a value of x, below which 90% of all
the observations of x lie. That implies, the median is the 50th percentile. Minimum value of a variable x is the 0th
percentile, and maximum value is the 100th percentile.
Quartiles, quintiles, deciles, etc. break the ordered set of observations into n equal parts: quartiles into 4 equal
parts corresponding to 25th, 50th, 75th percentiles; quintiles into five, deciles into ten.

Inter-quartile Range (IQR)


Difference between third (75th percentile) and first (25th percentile) quartiles. 𝐼𝑄𝑅 = 𝑄3 − 𝑄1

Linear Transforms of Data


Allows one to change the centre and spread of data and bring multiple variables onto a common point of reference
to allow for comparative analysis.

If transform is of the type 𝑌 = 𝑎 + 𝑏. 𝑋, then a represents a change of level and b represents a change of scale.
Some common effects of linear transforms on descriptive statistics:

average(a + bX) = a + b.average(X)


var(a + bX) = b2var(X)
median(a + bX) = a + b.median(X)
stdev(a + bX) = |b|.stdev(X)
IQR(a + bX) = |b|.IQR(x)

Z-score
A special (and important) type of linear transform of data. It is also called a standard score. Z-score of any
observation of a variable x, gives the number of standard deviations above or below the mean value that the
observation is located at.

𝑧 = (𝑥 − 𝜇)/𝜎

This transform maps the variable x to a distribution with mean 0 and variance 1, without altering the general
shape/characteristics of the original distribution. Essentially it recreates a pseudo-normal distribution for the
variable x.

Skewness and Kurtosis


Skewness measures the asymmetry of a distribution about the mean, and kurtosis measures the fatness of the
distribution.

Skewness = E(Z3). If skewness < 0, then left tail is longer; > 0, then right tail is longer.
Kurtosis = E(Z4) – 3. If kurtosis < 0, then narrow distribution; if > 0, then fat distribution.

Important Descriptive Statistics


1. Mean
2. Minimum
3. Maximum
4. Median
5. Inter-quartile range
6. Variance, or, standard deviation
7. Skewness
Normal Distribution

Special case: mean and median occur at same point; balanced distribution around mean.

Many people believe that datasets with very large number of observations will follow a normal distribution, or, be
close to a normal distribution – in my opinion, not necessary, and safer not to make this assumption.

Chebyshev’s Rule

Important takeaway: 95% of all observations lie within second standard deviation
Outliers
Classic rule: all observations for which Z > 2.

Preferred rule: using the boxplot, all observations that lie below Q1 – 1.5.IQR or above Q3 + 1.5.IQR

You might also like