SPSS practical (SLR)

Areas    Sales

1726.0  3681.0

1642.0  3895.0

2816.0  6653.0

5555.0  9543.0

1292.0  3418.0

2208.0  5563.0

1313.0  3660.0

1102.0  2694.0

3151.0  5468.0

1516.0  2898.0

5161.0  10674.0

4567.0  7585.0

5841.0  11760.0

3008.0  4085.0

·        First we plot the data as scatter plot

 

                     

·        Then we check correlations between them

Correlations

 

 

area

sales

area

Pearson Correlation

1

.954**

Sig. (2-tailed)

 

.000

N

14

14

sales

Pearson Correlation

.954**

1

Sig. (2-tailed)

.000

 

N

14

14

**. Correlation is significant at the 0.01 level (2-tailed).

 

 

 

 

 

 

 

 Correlation is sig. thus we can develop a linear between sales and area.

 

ANOVAb

Model

Sum of Squares

df

Mean Square

F

Sig.

1

Regression

1.062E8

1

1.062E8

121.009

.000a

Residual

1.053E7

12

877687.937

 

 

Total

1.167E8

13

 

 

 

a. Predictors: (Constant), area

 

 

 

 

b. Dependent Variable: sales

 

 

 

 

Fitted model is sig. significant amount of the variable of the responses (sales) variable has been captured by the fitted model. Sig. amount is given by R2

 

·           If the fitted model (SLR) is sig. does 2 parameters are significant???

                     NO

corr. is implies model is sig.

model is sig. implies beeta1 is sig.

model is sig. does not implies beeta0 is sig.

 

Coefficientsa

Model

Unstandardized Coefficients

Standardized Coefficients

t

Sig.

B

Std. Error

Beta

1

(Constant)

901.247

513.023

 

1.757

.104

area

1.686

.153

.954

11.000

.000

a. Dependent Variable: sales

 

 

 

 

Diagnostics error

Random –DM

Cvonst var –plot of yhat and error

Normality

DW statistic is close zero indicating error are random

 

Plot res vs predict indicates random confirming error having constant variance

 

Normality

 

 

Tests of Normality

 

Kolmogorov-Smirnova

Shapiro-Wilk

 

Statistic

df

Sig.

Statistic

df

Sig.

Unstandardized Residual

.137

14

.200*

.933

14

.333

a. Lilliefors Significance Correction

 

 

 

 

*. This is a lower bound of the true significance.

 

 

 

SW test stat is not sig. calming the errors are not significant different .

Mean is not sig. diff. from as zero is in 95 % CI .

Can you accept the model??

Percentage error = res*100/obs data

Percentage error varies -46% to 16%

Percentage error varies -19% to 16% with an exceptional one as -46%

 

SPSS practical (MLR)

Y

X1

X2

X3

X4

27

20

50

75

15

23

27

55

60

20

18

22

62

68

16

26

27

55

60

20

23

24

75

72

8

27

30

62

73

18

30

32

79

71

11

23

24

75

72

8

22

22

62

68

16

24

27

55

60

20

16

40

90

78

32

28

32

79

71

11

31

50

84

72

12

22

40

90

78

32

24

20

50

75

15

31

50

84

72

12

29

30

62

73

18

22

27

55

60

20

 

 

Correlations matrix among Y & explanatory variables

 

 

y

x1

x2

x3

x4

y

Pearson Correlation

1

.373

.059

.048

-.522*

Sig. (2-tailed)

 

.127

.815

.852

.026

x1

Pearson Correlation

.373

1

.758**

.288

.192

Sig. (2-tailed)

.127

 

.000

.247

.444

x2

Pearson Correlation

.059

.758**

1

.555*

.099

Sig. (2-tailed)

.815

.000

 

.017

.697

x3

Pearson Correlation

.048

.288

.555*

1

.060

Sig. (2-tailed)

.852

.247

.017

 

.813

x4

Pearson Correlation

-.522*

.192

.099

.060

1

Sig. (2-tailed)

.026

.444

.697

.813

 

*. Correlation is significant at the 0.05 level (2-tailed).

 

 

**. Correlation is significant at the 0.01 level (2-tailed).

 

 

Y is sig. corr. only with x4.

Though explanatory variables supposed to be independent, x1 & x2 as well as x2 and x3 are significant.

 

Coefficientsa

Model

Unstandardized Coefficients

Standardized Coefficients

t

Sig.

B

Std. Error

Beta

1

(Constant)

22.625

7.819

 

2.894

.013

x1

.468

.114

1.024

4.112

.001

x2

-.235

.086

-.777

-2.747

.017

x3

.156

.134

.224

1.166

.265

x4

-.407

.098

-.656

-4.150

.001

a. Dependent Variable: y

 

 

 

 

In SLR model is sig. implies variable is sig.

But MLR model sig. does not implies all variables are sig.

Adding models increases the R2, but it does not imply all variable are sig.

Non sig. variables are known nuisance variable.

Prob: how do we decide which parameter to be included in to the model?