Statistics

1. Range = Maximum value – minimum value

𝑄3 − 𝑄1
2. Quartile deviation = 2
𝑛+1
(a) For discrete data : Q1 = th observation
4
3(𝑛 + 1)
Q3 = th observation
4
(b) For continuous frequency distribution
𝑛
− 𝐶
Q1 = l + 4 ×h
𝑓
3𝑛
− 𝐶
4
Q3 = l + ×h
𝑓
3. Mean deviation
(a) Mean deviation for ungrouped data
∑|(𝑥𝑖 − 𝑋̅ )|
M.D.(𝑋̅) = (about mean)
𝑛
∑|𝑥𝑖 − 𝑀|
M.D.(𝑀) = (about median)
𝑛

(b) Mean deviation for discrete frequency distribution

∑ 𝑓 |𝑥 − 𝑋| ̅
M.D.(𝑋̅) = 𝑖∑ 𝑓𝑖 (about mean)
𝑖
∑ 𝑓𝑖 |𝑥𝑖 − 𝑀|
M.D.(𝑀) = ∑ 𝑓𝑖
(about median)

(c) Mean deviation for continuous frequency distribution

∑ 𝑓 |𝑥 − 𝑋| ̅
M.D.(𝑋̅) = 𝑖∑ 𝑖 (about mean)
𝑓𝑖
∑ 𝑓𝑖 |𝑥𝑖 − 𝑀|
M.D.(𝑀) = ∑ 𝑓𝑖
(about median)

4. Variance and Deviation

∑(𝑥𝑖 − 𝑋̅ )2
(i) 𝜎 2 or Var(x) = (for ungrouped data)
𝑛

∑ 𝑓𝑖 (𝑥𝑖 − 𝑋̅)2 ∑ 𝑓𝑖 𝑥𝑖2 ∑ 𝑓 𝑖 𝑥𝑖 2

(ii) 𝜎 2 or Var(x) = ∑ 𝑓𝑖
or ∑ 𝑓𝑖
− ( ∑ 𝑓𝑖
) (for grouped data)

∑ 𝑓𝑖 𝑡𝑖2 ∑𝑓 𝑡 2 𝑥𝑖 − 𝐴
(iii) 𝜎 2 or Var(x) = 𝑐 2 [ ∑ 𝑓𝑖
− ( ∑ 𝑓𝑖 𝑖 ) ] where 𝑡𝑖 = (Step deviation method)
𝑖 𝑐

Deviation = 𝜎 = √𝑆. 𝐷.
5. Moments
∑(𝑥𝑖 − 𝑋̅)𝑟
(i) For ungrouped data : 𝜇𝑟 = where r = 0, 1, 2,……
𝑛

∑ 𝑓𝑖 (𝑥𝑖 − 𝑋̅)𝑟
(ii) For grouped data : 𝜇𝑟 = where r = 0, 1, 2,……
𝑛

6. Absolute measures of skewness

(i) Absolute skewness = Q3 + Q1 – 2Q2
(ii) If Q3 – Q2 > Q2 – Q1, then skewness is positive
(iii) If Q3 – Q2 < Q2 – Q1, then skewness is negative
7. Relative measures of skewness
(i) Karl Pearson’s Coefficient of Skewness:
𝑀𝑒𝑎𝑛−𝑀𝑜𝑑𝑒
SKp = 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛

If mode is ill defined, then

3(𝑀𝑒𝑎𝑛−𝑀𝑒𝑑𝑖𝑎𝑛)
SKp = 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛

(ii) Bowley’s Coefficient of Skewness

𝑄3 + 𝑄1 −2𝑀𝑒𝑑𝑖𝑎𝑛 𝑄3 + 𝑄1 −2𝑄2
SKB = =
𝑄3 − 𝑄1 𝑄3 − 𝑄1

If SKB > 0, then distribution is positively skewed.

If SKB = 0, then distribution is symmetrical.
If SKB > 0, then distribution is negatively skewed.
(iii) Moment coefficient of Skewness
𝜇3
𝛾1 = ±√𝛽1 =
√𝜇23

Generally, skewness is measured in terms of

𝜇2
𝛽1 = 𝜇33 where sign of skewness is determined by sign of 𝜇3 .
2

8. KURTOSIS
Moment Coefficient of Kurtosis
𝜇
𝛽2 = 𝜇42
2

9. Covariance of X and Y
∑(𝑥− 𝑋̅ )(𝑦− 𝑌̅) 1 1
Cov(X, Y) or CXY = or [∑ 𝑥𝑦 − ∑ 𝑥 ∑ 𝑦]
𝑁 𝑁 𝑁

1 1
Or Cov(X, Y) or CXY = [∑ 𝑢𝑣 − ∑ 𝑢 ∑ 𝑣] where u = x – A, v = y – B
𝑁 𝑁

10. Karl Pearson’s Coefficient of Correlation

𝐶𝑜𝑣(𝑥,𝑦) 𝐶𝑜𝑣(𝑥,𝑦)
r or 𝜌(x, y) = =
√𝑉𝑎𝑟(𝑥)√𝑉𝑎𝑟(𝑦) 𝜎𝑥 𝜎𝑦

(i) If x - 𝑋̅, y - 𝑌̅ are small fractionless numbers, we use

∑(𝑥− 𝑋̅)(𝑦− 𝑌̅)
r =
√∑(𝑥𝑖 − 𝑋̅)2 √∑(𝑦𝑖 − 𝑌̅)2
(ii) If x, y are small numbers, we use
1
∑ 𝑥𝑦− ∑𝑥∑𝑦
𝑁
r= 1 1
√∑ 𝑥 2 − (∑ 𝑥)2 √∑ 𝑦 2 − (∑ 𝑦)2
𝑁 𝑁
 (iii) If we use assumed mean A and B,
u = x – A, v = y – B
1
∑ 𝑢𝑣− ∑𝑢∑𝑣
𝑁
r= 1 1
√∑ 𝑢2 − (∑ 𝑢)2 √∑ 𝑣 2 − (∑ 𝑣)2
𝑁 𝑁

11. Spearman’s Rank Correlation Coefficient

6 ∑ 𝑑2
r = 1 - 𝑛(𝑛2−1)

where d = difference between ranks of corresponding x and y

n = number of pairs of values (x, y) in the data
12. Percentile Rank
(i) Percentile rank for ungrouped data
𝐿+0.5𝐸
PR = × 100
𝑛

where L = number of observations (scores) less than xi

E = number of observations (scores) equal to xi
n = total number of observations (scores)
(ii) Percentile rank for grouped data (discontinuous)
𝐶+0.5𝑓
PR = × 100
𝑛

where C = cumulative frequency preceding the score xi

f = frequency of score xi
n = sum of all frequencies
(iii) Percentile rank for grouped data (continuous)
𝑥 100
PR = (𝐶 + × 𝑓) ×
ℎ 𝑛

where C = cumulative frequency preceding the percentile class

f = frequency of the percentile class
x = score whose percentile rank is required – lower limit of percentile class
h = length of class interval
n = sum of all frequencies