REGRESSION

Regression converts data clumps into comprehensible information.

Understanding the impact of:

the INDEPENDENT CAUSAL VARIABLE REGRESSOR (X) on
the concerned DEPENDENT AFFECTED VARIABLE REGRESSAND (Y)
Y = f (X)

All economic phenomena has to be understood in terms of

a set of CAUSAL VARIABLES (X1, X2, X3, …, XN)
impacting the DEPENDENT VARIABLE (Y)

Y = f (X1, X2, … XN)

 Regression (contd.)

INDEPENDENT VARIABLE (X) DEPENDENT VARIABLE (Y)

PREDETERMINED Variable DETERMINED Variable

It serves as: It serves as:

the CAUSE the EFFECT
or the STIMULANT or the CONSEQUENCE

Y is also called a STOCHASTIC Variable since ....

 Regression (contd.)

This functional relationship Y = f (X) could either be:

Positive / Direct Negative / Indirect / Inverse

X Y X Y

X Y
X Y
Y
Y

EFFECT
EFFECT

CAUSE X
CAUSE X

Slope: Positive Slope: Negative

Trajectory: UPWARD SLOPING Trajectory: DOWNWARD SLOPING
 Regression (contd.)

Data can be converted into an equation, which will indicate the curve’s:

TRAJECTORY and VALUE OF THE SLOPE

Via Regression data can be converted into a LINEAR equation

Y = mX + h OR Y = h + mX PRF

SRF

What is the significance of ‘h’ ???

 Regression (contd.) OLS Method

X Y

1 17 -2 -3 6 4 21 -4 16 9
2 20 -1 0 0 1 20.5 -0.5 0.25 0
3 27 0 7 0 0 20 7 49 49
4 23 1 3 3 1 19.5 3.5 12.25 9
5 13 2 -7 -14 4 19 -6 36 49
15 100 -5 10 113.5 116
 Regression (contd.)

1 2 3 4 5 X
 Regression (contd.)

(R2) Coefficient of Determination of Goodness of Fit

For the given data R2 = 0.02 or 2%

Hence, X cannot explain 98 % of Y’s behaviour

 Regression (contd.)

X Y
10 06 -22 -16 352 484 256

21 31 -11 9 -99 121 81

32 21 0 -1 0 0 1

45 12 13 -10 -130 169 100

52 40 20 18 360 400 324

160 110 483 1174 762

 Regression (contd.)

Illustration for FITTING OF STRAIGHT-LINE TREND: (ODD No. of years)

Given the data, fit a trend equation Year (X’) 2000 2001 2002 2003 2004 2005 2006
Estimate demand for 2008 and 1998 Demand (Y) 30 35 40 42 45 48 50
(Rs ‘00000)
Solution:

Year Demand (Y) X = (Year – 2003) XY X2 Trend values

2000 30 -3 -90 9 32
2001 35 -2 -70 4 35
2002 40 -1 -40 1 38
2003 42 0 0 0 41
2004 45 1 45 1 45
2005 48 2 96 4 48
2006 50 3 150 9 51
Total 290 0 91 28

Estimated Demand for: 2008 41.43 + 3.25(5) = 57.68 1998 41.43 + 3.25(-5) = 25.18
 Regression (contd.)

Illustration for FITTING OF STRAIGHT-LINE TREND: (EVEN No. of Years)

Fit a trend equation for the following data and estimate Demand for 2009 and 1998

Solution:
Year Demand (Y) X = (Year – 2003.5) XY X2 Trend values
2000 30 -3.5 -105 12.25 30.75
2001 35 -2.5 -87.5 6.25 34.46
2002 40 -1.5 -60 2.25 38.17
2003 42 -0.5 -21 0.25 41.89
2004 45 0.5 22.5 0.25 45.6
2005 48 1.5 72 2.25 49.3
2006 50 2.5 125 6.25 53.02
2007 60 3.5 210 12.25 56.74
Total 350 0 156 42

Estimated demand for 2009: Estimated demand for 1998:

43.75 + 3.71(5.5) = 64.15 43.75 + 3.71(-5.5) = 23.35
 Regression (contd.)

FITTING AN EXPONENTIAL (= non-linear) TREND :

Exponential trend is applicable where growth in a time series data
is at a constant rate with respect to time.
The exponential curve is given by the equation:

This equation has to be transformed into a log-linear form as:

and
 Regression (contd.)

FITTING AN EXPONENTIAL TREND (Contd.): (ODD Number of Years)

Illustration:

The indices of the market demand for Good X from 1975 to 1985 are
presented below. Find the exponential trend equation and the estimated
demand for 1988

Year ‘75 ‘76 ‘77 ‘78 ‘79 ‘80 ‘81 ‘82 ‘83 ‘84 ‘85
Demand 100 115 130 137 135 130 140 148 155 162 180
 Regression (contd.)

FITTING AN EXPONENTIAL TREND (Contd.): (ODD Number of Years)

Solution:
Year Demand (Y) X log Y [X (log Y)] X2
1975 100 -5 2.0000 -10 25
1976 115 -4 2.0607 -8.2428 16
1977 130 -3 2.1139 -6.3417 9
1978 137 -2 2.1367 -4.2734 4
1979 135 -1 2.1303 -2.1303 1
1980 130 0 2.1139 0 0
1981 140 1 2.1461 2.1461 1
1982 148 2 2.1703 4.3406 4
1983 155 3 2.1903 6.5709 9
1984 162 4 2.2095 8.838 16 Antilog (2.14) = 138.03
1985 180 5 2.2553 11.2765 25 Antilog (0.02) = 1.05
Totals 23.5272 2.1839 110
 Regression (contd.)

FITTING AN EXPONENTIAL TREND (Contd.): (ODD Number of Years)

 Regression (contd.)

FITTING AN EXPONENTIAL TREND (Contd.): (EVEN Number of Years)

Illustration:

The indices of demand for Good Y in India from 1975 to 1986 are presented
below. Find the exponential trend equation and the estimated index for 1988

Year ‘75 ‘76 ‘77 ‘78 ‘79 ‘80 ‘81 ‘82 ‘83 ‘84 ‘85 ‘86
Index 100 115 130 137 135 130 140 148 155 162 180 170
 Regression (contd.)
FITTING AN EXPONENTIAL TREND (Contd.): (EVEN Number of Years)

Solution:
Demand
Year
(Y)
log Y X [X (log Y)] X2
1975 100 2.0000 -5.5 -11 30.25 68.10
1976 115 2.0607 -4.5 -9.2732 20.25 77.63
1977 130 2.1139 -3.5 -7.3987 12.25 88.50
1978 137 2.1367 -2.5 -5.3418 6.25 100.89
1979 135 2.1303 -1.5 -3.1955 2.25 115.02
1980 130 2.1139 -0.5 -1.0570 0.25 131.12
1981 140 2.1461 0.5 1.0731 0.25 149.48
1982 148 2.1703 1.5 3.2555 2.25 170.41
1983 155 2.1903 2.5 5.4758 6.25 194.26
2.2095 3.5 7.7333 12.25 221.46 Antilog (2.1465) = 140.12
1984 162
2.2553 4.5 10.1489 20.25 252.46 Antilog (0.0188) = 1.0442
1985 180
1986 170 2.2305 5.5 12.2678 30.25 287.81
Totals 25.7577 2.6882 143
CONCAVITY and CONVEXITY
 Concave to X-Axis Downwards Convex to X-Axis Downwards

Y Y
For more of X we give up For more of X we give up
more & more of Y Less & less of Y

X X
 Concave to X-Axis Upwards Convex to X-Axis Upwards

Y Y For more of X we get

For more of X we get
less & less addition of Y more & more addition of Y

X
X
 Convex to X-Axis Upwards Convex to X-Axis Downwards

Y Y

For more of X we get For more of X we give

more & more of Y
up
less & less of Y

X X
 Concave to X-Axis Upwards Concave to X-Axis Downwards

Y For more of X we get Y give up

For more of X we
more & more of Y
less & less increment in Y

X X
LEARNING CURVE
 Learning Curve
A learning curve alludes to the time it takes for an employee to learn the
process or system that is involved in the production process

It is a visual representation of the relationship between how proficient an

individual is at a task and the amount of experience they have in this job

The learning curve is based on the premise that individuals require time to
become proficient at something new

For a business, this means that investment needs to be made in training

personnel to obtain a certain output.

Over time, the individual trainee will learn and become more efficient at that
task and so increase productivity

Thus, when this concept is applied to business, it indicates to the relationship

between cost and output
 Learning Curve (Contd.)

A Typical Learning Curve

Performance measure
Plateau Phase

Fast-paced Phase

Slow-paced Phase
Number of attempts at learning

A Typical Learning Curve of a A Typical Learning Curve of a A Typical Learning Curve when
GENIUS NON-GENIUS loss of learning capacity sets in
Performance measure
Performance measure

Number of attempts at learning Performance measure

Number of attempts at learning Number of attempts at learning
 Learning Curve (Contd.)

The following are the types of learning curves:

Diminishing Returns LC: It is used to illustrate tasks that are quick to learn
and early to plateau; learning manual tasks tend to be in this category

Increasing Returns LC: It is used to signify tasks that are difficult to learn at
first and where the rate of returns are significant after some time; learning
to operate a sophisticated instrument

Sigmoid or S-LC: It is also known as the increasing – decreasing return

curve - it represents a task that may be difficult for an individual to learn
initially but once the individual becomes proficient, it begins to plateau

Complex LC: It the learning trajectory traced over a long period of time - in
this curve, the individual may experience a temporary belief of mastery, only
to uncover that s/he has even more to learn
INFLEXION
 Inflexion

A 6 B 4 C 2
A 6 C 2

Points of INFLEXION C 2

B 4
Model 3 A 6

Model 4 Model 5 Model 6 B 4

D
A 6 C 2

E G Model 1 Model 2
H

No INFLEXION Points
Model 7 Model 8
F