Step 1:

Identify any three variables (minimum), preferably time series variable, which
you think have a sound theoretical relationship among themselves in real
life(preferably some contemporary issues in Business). Identify one of them as
dependent variable(Y) and others as independent variable (X’s).
Step 2:
Collect the data pertaining to each variables relating to any country context,
preferable the data and variables from your own company, if it is allowed. The
number of observation for each data set should not be less than 20 observations.
More is preferable and there is no upper limit for the data points and number of

I have taken my co Sterling Commerce PAT of last 4 years with 2 Independent

variable
Time  For each quarter from 2005 to 2009
Dependent variable PAT Profit After Tax in $Million
Independent variable Employee CTC (Salary) in $Million
Independent variable Miscellaneous Expense in $Million

Misc
Year Quarter t(x) PAT(y) Employee CTC Expenses
2005 1 1 6 7 9
  2 2 5 8 12
  3 3 7 10 14
  4 4 7 12 14
2006 1 5 7 12 13
  2 6 8 13 15
  3 7 6 13 18
  4 8 9 14 14
2007 1 9 8 15 20
  2 10 10 16 14
  3 11 7 16 17
  4 12 9 17 20
2008 1 13 8 18 18
  2 14 12 19 22
  3 15 11 20 25
  4 16 10 20 22
2009 1 17 12 20 18
  2 18 14 18 19
  3 19 13 18 16
  4 20 14 18 17
Part-II

0 A)Plot your data (any one variable which you are interested to
forecast) against time

Time Vs PAT

0
1
2 b) Determining trend, cyclical, or seasonal or random
pattern.
Using Decomposition

Trend Projection
Eq of trend line of best fit
y= a + bx

b=∑x y- n x` y` / ∑ x² - n (x`)²

Year Quarter t(x) PAT(y) Employee CTC Misc Expenss x² xy

2005 1 1 6 7 9 1 6
  2 2 5 8 12 4 10
  3 3 7 10 14 9 21
  4 4 7 12 14 16 28
2006 1 5 7 12 13 25 35
  2 6 8 13 15 36 48
  3 7 6 13 18 49 42
  4 8 9 14 14 64 72
2007 1 9 8 15 20 81 72
  2 10 10 16 14 100 100
   3 11 7 16 17 121 77
  4 12 9 17 20 144 108
2008 1 13 8 18 18 169 104
  2 14 12 19 22 196 168
  3 15 11 20 25 225 165
  4 16 10 20 22 256 160
2009 1 17 12 20 18 289 204
  2 18 14 18 19 324 252
  3 19 13 18 16 361 247
  4 20 14 18 17 400 280

∑ 210 183 2870 2199

x` = ∑x/n = 210/20 = 10.5

y` = ∑y/n = 183/20 = 9.15

b = (2199 – 20 * 10.5 * 9.15) / (2870 – 20 * 10.5²)

b = 4.77

a = y` - b x` = 9.15 – 4.77 * 10.5 = .42

Trend line eq y = 4.77 + .42x

Calculating Trend T using above formula

Calculation of Seasonal variation S

Qtr I Qtr II Qtr II Qtr IV

2005 6 5 7 7
2006 7 8 6 9
2007 8 10 7 9
2008 8 12 11 10
2009 12 14 13 14

∑ 41 49 44 49
Avg=∑/ 5 8.2 9.8 8.8 9.8
S Index=Avg / 1.07103
(∑Avg/4) 0.896175 8 0.961748634 1.071038251

Avg of Qtr Avg = 8.2+9.8+8.8+9.8 / 4 = 9.15

Qtr 1 S Index = 8.2/9.15 = 0.896175

After Finding T and S Cyclical & Irregulaitis Variation can be calculated by C.I = T.S.C.I / T.S
Cyclical variation can be calculated by three period moving avg
For 2005 2nd Qtr = 1.289968 + 0.832181 +1.207094 / 3 = 1.109748 and so on for other period
Once Cyclical component determine then I can be calculated by = T.S.C.I / T.S.C

C
Three
CI period I
Employee Misc (T.S.C.I / moving T.S.C.I
Year Quarter t(x) PAT(y) CTC Expenses T S T.S) avg / T.S.C
2005 1 1 6 7 9 5.19 0.8962 1.289968    
  2 2 5 8 12 5.61 1.071 0.832181 1.109748 0.74988
  3 3 7 10 14 6.03 0.9617 1.207094 1.017533 1.18629
  4 4 7 12 14 6.45 1.071 1.013325 1.119119 0.90546
2006 1 5 7 12 13 6.87 0.8962 1.136937 1.058302 1.07430
  2 6 8 13 15 7.29 1.071 1.024644 0.990261 1.03472
  3 7 6 13 18 7.71 0.9617 0.809203 0.955823 0.84660
  4 8 9 14 14 8.13 1.071 1.033624 0.96229 1.07412
2007 1 9 8 15 20 8.55 0.8962 1.044044 1.03953 1.00434
  2 10 10 16 14 8.97 1.071 1.040922 0.953376 1.09182
  3 11 7 16 17 9.39 0.9617 0.775163 0.890899 0.87009
  4 12 9 17 20 9.81 1.071 0.856612 0.834788 1.02614
2008 1 13 8 18 18 10.23 0.8962 0.872588 0.927088 0.94121
  2 14 12 19 22 10.65 1.071 1.052064 0.985967 1.06703
  3 15 11 20 25 11.07 0.9617 1.03325 0.96598 1.06963
  4 16 10 20 22 11.49 1.071 0.812626 0.990043 0.82079
2009 1 17 12 20 18 11.91 0.8962 1.124254 0.999017 1.12536
  2 18 14 18 19 12.33 1.071 1.06017 1.081546 0.98023
  3 19 13 18 16 12.75 0.9617 1.060214 1.037645 1.0217
  4 20 14 18 17 13.17 1.071 0.992551    

As calculated above there is Trend and Seasonal component but Cyclical and
irregularities and near 1.0 means C and I has negligible effect

Calculating Forecast for 2010 Qtr 1 as follows

Trend = 4.77 + .42 * 21 =13.6
Forecast for period 21 = 13.6 * .8962 * 1.0 * 1.0=12.18

C & D)Naïve_data Forecast-a gives the best result.Aslo Holt’s and winter’s method
gives the good result MSE was very low for them.Pl see the attached time-series xls
sheet
 time-series.xls

Regression Model Part(1)

1 A)Plot the data in a scatter form (for two variable cases) and determine what kind of
curve (model) you will fit for the data set.
2 Scatter data between PAT and Annual Salary

25

20

15
Series1
10

0
0 2 4 6 8 10 12 14 16

3 Scatter data between PAT and Misc Expense

30

25

20

15 Series2

10

0
0 2 4 6 8 10 12 14 16
1 B) Obtain the intercept and slope coefficients and interpret the regression results

Employee Misc
PAT(x) CTC(y) Expenss(z) x² y² z² xy xz yz
6 7 9 36 49 81 42 54 63
5 8 12 25 64 144 40 60 96
7 10 14 49 100 196 70 98 140
7 12 14 49 144 196 84 98 168
7 12 13 49 144 169 84 91 156
8 13 15 64 169 225 104 120 195
6 13 18 36 169 324 78 108 234
9 14 14 81 196 196 126 126 196
8 15 20 64 225 400 120 160 300
10 16 14 100 256 196 160 140 224
7 16 17 49 256 289 112 119 272
9 17 20 81 289 400 153 180 340
8 18 18 64 324 324 144 144 324
12 19 22 144 361 484 228 264 418
11 20 25 121 400 625 220 275 500
10 20 22 100 400 484 200 220 440
12 20 18 144 400 324 240 216 360
14 18 19 196 324 361 252 266 342
13 18 16 169 324 256 234 208 288
14 18 17 196 324 289 252 238 306
183 304 337 1817 4918 5963 2943 3185 5362
2

x=Dependent variable PAT

y=Employee salary
z=Misc Expenses
Least square Normal eq
x=a+by+cz

∑x=na+b∑y+c∑z

183=20a+304b+337c ------- eq 1

∑xy=a∑y +b∑y²+c∑yz
2943=304a+4918b+5362c ------- eq 2

∑xz=a∑z +b∑yz+c∑z²

3185=337a+5362b+5963c ------- eq 3

Multiply eq1 by 304 and eq 2 by 20

55632=6080a+92416b+102448c
58860=6080a +98360b+107240c
3228=5944b+4792c ---------------eq 4

Multiply eq1 by 337 and eq 3 by 20

61671=6740a+102448b+113569c
63700=6740a +107240b+119260c
2029=4792b+5691c -------------- eq 5

Multiply eq1 by 5691 and eq 2 by 4792

18370548=33827304b+27271272c

9722968= 22963264b+27271272c
8647580 =10864040b

b = .8

Substituting b in eq 4

c =- -.32

Substituting b and c in eq 1

183=20a+304 * .8 + 337 * (-.32)

a =2.38

Least Square regression Give by

 x=2.32 + .8y - .32z

intercept a= 2.32
Mean In absence of all variable there will be PAT of 2.32 million
Doller

slope coefficients b=.8

Mean Every unit increase in Employee sal will accounted for an
increase of .8 in value of PAT
slope coefficients b=-.32
Mean Every unit increase in Misc Expense will accounted for an
decrese of .32 in value of PAT

1 C) Apply‘t’ test and ‘F’ test to test the hypothesis of the parameters with appropriate
significance level
Employee Misc
PAT(x) CTC(y) Expenss(z) x` (x`- x~)² (x-x`)²
6 7 9 5.04 16.8921 0.9216
5 8 12 4.88 18.2329 0.0144
7 10 14 5.84 10.9561 1.3456
7 12 14 7.44 2.9241 0.1936
7 12 13 7.76 1.9321 0.5776
8 13 15 7.92 1.5129 0.0064
6 13 18 6.96 4.7961 0.9216
9 14 14 9.04 0.0121 0.0016
8 15 20 7.92 1.5129 0.0064
10 16 14 10.64 2.2201 0.4096
7 16 17 9.68 0.2809 7.1824
9 17 20 9.52 0.1369 0.2704
8 18 18 10.96 3.2761 8.7616
12 19 22 10.48 1.7689 2.3104
11 20 25 10.32 1.3689 0.4624
10 20 22 11.28 4.5369 1.6384
12 20 18 12.56 11.6281 0.3136
14 18 19 10.64 2.2201 11.2896
13 18 16 11.6 6.0025 1.96
14 18 17 11.28 4.5369 7.3984
183       96.7476 45.9856
Mean
x~=9.15          
2
Determine SSR = ∑(x`-x~)²=96.74
Determine SSE = ∑ (x-x`)²=45.99
k=No of independent variable
MSR = SSR/k = 96.74/2=48.36
 MSE = SSE/n-(k+1) = 45.99/20-(2+1) =2.71

F ratio = MSR/MSE= 48.36/2.71=17.84

Table value of F for SSR degreee of freedom k = 2 and SSE degree of freedom n-
(k+1) = 17 and α=.05 is 3.59 .Since calculated value of F = 17.84 is higher then table
value the null hypothesis is rejected.
Hence we conclude that regression model has no significance prediction for the
dependent variable in other word rejection of H0 mean at least one of the
independent variable is adding significant prediction for PAT.

T test t=√F test

t=4.2
Table value of t for degreee of freedom 18 and α=.05 is 2.1 .Since calculated value
of t = 4.2 is higher then table value the null hypothesis is rejected.
Hence we conclude that regression model has no significance prediction for the
dependent variable in other word rejection of H0 mean at least one of the
independent variable is adding significant prediction for PAT.

1 d) Obtain the forecasted value from the model for some fixed values of independent
variable.
Employee Misc
CTC(y) Expenss(z) Forecast PAT
20 20 11.92
22 18 14.16
24 14 17.04
28 22 17.68
28 28 15.76

As shown above if Employee CTC(Adding more employee) will increase then PAT will
increase but if Misc expense will increase then PAT will decrease

Part III Give your own comments on the above results and provide suggestion to the
company.

Time series analysis comprises methods for analyzing time series data in
order to extract meaningful statistics and other characteristics of
the data. This useful to identify the pattern in data .but it not
suitable for finding the more than one independent variable and
its effect on PAT. To identify pattern in PAT for any seasonality
or cyclic or irregularities effect this is useful as shown above that
 16

14

12

10
Actual
8
Holt's Forecast
6

0
1 2 3 4 5 6 7 8 9

16

14

12

10
Actual
8
Winter
6

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

regression analysis includes any techniques for modeling and analyzing

several variables, when the focus is on the relationship between
a dependent variable and one or more independent variables.It
does not indicate any seasonality or or cyclic or irregularities
effect.Actual data and forecasted data are very near with this as
shown below .
 16

14

12

10
Actual
8
Regression Forecast
6

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

In my case both time series and regression analysis able to give forecasted
value well in range.