CG ver:1.

0 date:20201116 Statistics I Page 1 of 3

VARIANCE
(x − μ)2

∑
n u m e r a t or =
2
∑ (x − μ)2 2
∑ (x − x̄)2
σpopulation = Ssample = =
∑
(x 2 − 2x μ + μ 2 )

N n−1 (x 2 ) − 2μ μ 2

∑ ∑ ∑
= x+
(x 2 ) − 2μ . n . μ + n μ 2

∑
=
Sum of squared deviations of every observation from its mean over n
total number of observations. Units are square units of the original (x 2 ) − n μ 2

∑ i
=
variable.
1
Computational Formulae :

n n
∑1 (xi2) − Nμ 2 ∑1 (xi2) − n x̄ 2 <= useful for manual calculations
2 2
σpopulation = Ssample =
N n−1
STANDARD DEVIATION
Square root of Variance. Units are associated with numerical measures. σ or s

PERCENTILES, QUARTILES, IQR,Box Plot

100p percent of data ≤ percentile ≤ 100(1-p) percent of data.

Compute : Arrange ascending, decimal np? next pos : integer np? avg of value at np&np+1.
Percentile need not be part of data set. Q1 = 25th p, Q2= median = 50th p, Q3= 75th p.
5 number summary : Min,Q1,Q2,Q3,Max

IQR = Q3-Q1 ; Range = Max-Min

OUTLIERS
outlier < Q1 − 1.5 * IQ R; outlier > Q3 + 1.5 * IQ R
TWO WAY CONTINGENCY TABLE
Association between 2 categorical variables - relative frequencies. Bivariate Categorical Data.

Row Total and Column Total. Row Relative Frequencies (cell frequency/row total) & Column
Relative Frequencies (cell frequency/column total). If the row or column relative frequencies are
the same for all rows/columns then the 2 variables are not associated.

STACKED/SEGMENTED BAR CHART

Represents counts of a particular category and segments representing frequency of category. 100
percent stacked bar chart is useful for point to whole relationships. (e.g. Gender vs Phone
ownership)

SCATTER PLOT
y-axis : response variable ; x-axis : explanatory variable

e.g. x-axis : age and y-axis : height


CG ver:1.0 date:20201116 Statistics I Page 2 of 3
COVARIANCE

n n ∑n1 xi . ∑n1 yi
∑1 (xi − x̄)(yi − ȳ) ∑1 (xi . yi ) − N
cov(x, y)population = =

N N
n

∑
num erator = (xi − x̄)(yi − ȳ)

1
n

∑
= (xi . yi − xi . ȳ − x̄ . yi + x̄ . ȳ)

1
n n n n

∑ ∑ ∑ ∑
= xi . yi − xi . ȳ − x̄ . yi + x̄ . ȳ

1 1 1 1
n n n n

∑ ∑ ∑ ∑
= xi . yi − ȳ . xi − x̄ . yi + x̄ . ȳ

1 1 1 1
n

∑
= xi . yi − ȳ . N . x̄ − x̄ . N . ȳ + N . x̄ . ȳ

1
n

∑
= xi . yi − N . x̄ . ȳ

n n
n ∑1 xi . ∑1 yi
∑
= xi . yi −

1
N

n ∑n1 xi . ∑n1 yi n ∑n1 xi . ∑n1 yi

∑1 xi . yi − N
∑1 xi . yi − n
cov(x, y)population = cov(x, y)sample =

N n−1
Covariance quantifies the strength of the linear association between 2 numerical variables. Units
of Covariance: x-variable X y-variable (e.g. years.cm or years.Rs); hence difficult to interpret.
When both variables are moving in the same direction covariance is positive.

CORRELATION COEFFICIENT

n n ∑n1 xi . ∑n1 yi
cov(x, y) ∑1 (xi − x̄)(yi − ȳ) ∑1 xi . yi − n
r = = =

Sx . Sy n
∑1 (xi − x̄)2 .
n
∑1 (yi − ȳ)2 ( ∑ x 2 − n . x̄2) . ( ∑ y 2 − n . ȳ2)

−1 ≤ r ≤ + 1
The Pearson correlation coefficient is derived from covariance. Divide covariance of x and y by
product of standard deviations of x and y. Units of standard deviations cancel out the units of
covariance.

CG ver:1.0 date:20201116 Statistics I Page 3 of 3

POINT BI-SERIAL CORRELATION COEFFICIENT

For a dichotomous categorical variable (2 categories) . X is a numerical variable and Y is a

categorical variable.

e.g.: Gender(Y0 and Y1) and Marks(X)

( Sx )
Ȳ0 − Ȳ1 n0 n
rpb = . p0 . p1 p0 = ; p1 = 1
n n

GOODNESS OF FIT

R2 = r 2

R 2 is a measure of the proportion of variance in the data-set explained by the explanatory

variable. The measure is closer to 1 when r is closer to -1 or +1.

EFFECT OF MANIPULATING DATA WITH CONSTANT

Add Constant (+C) Multiply Constant (xC) Outliers

Mean +C *C Affected

Median +C *C Not Affected

Mode +C *C Not Affected

Variance no effect * C2 Affected

Standard Deviation no effect *C Affected

Covariance ? ? ?

Correlation Coefficient ? ? ?

Calculation Tables for Computational Formulae (Shortcuts)

n
Variance (sample) x x2 ∑1 (xi2 ) − n x̄ 2
2
Ssample =
n−1
Covariance (sample) x y xy n ∑n1 xi . ∑n1 yi
∑1 xi . yi − n
=
n−1
∑n1 xi . ∑n1 yi
Correlation Coefficient x y xy x2 y2 n
∑1 xi . yi − n
=
( ∑ x2 −n . x̄2) . ( ∑ y 2 − n . ȳ2)