STATS 362

Statistical Computing and Data Analysis 2

 UNIT 1
• REGRESSION

1. LINEAR REGRESSION MODEL

2. MULTIPLE REGRESSION MODEL

3. EXPONENTIAL REGRESSION

1
 REGRESSION
• FORMS OF REGRESSION

▪ LINEAR
▪ QUADRATIC
▪ CUBIC
▪ QUARTIC

• MULTIPLE REGRESSION
• EXPONENTIAL REGRESSION
2
 LINEAR REGRESSION
A major objective of many statistical investigations is to establish
relationships that make it possible to predict one or more variables in
terms of others. This regression is of the form y=ax+b, where
Y is the dependent variable.
X is the independent variable.
a is the y-intercept.
And b is the coefficient.
Note: The command ‘lm()’ is used to perform least squares regression.

3
R-code for linear regression
>x=c(numeric values)
>y=c(numeric values)
>a=lm(y~x)
>summary(a)

• Here y is the dependent variable

• X is the independent variable

4
 EXAMPLE
The following table presents data on the fretting wear of mild steel and oil
viscosity. Where, x=oil viscosity and y =wear volume.

Viscosity(X) 3.9 2.1 6.4 5.7 4.7 2.8 3.4 7.5 3 4.5
Volume(Y) 6.5 7.8 5.2 8.2 9.2 8.9 7.3 9.8 5.6 4.6

Using the R software, fit an appropriate linear, quadratic, cubic and quartic
model to these data.

5
 SOLUTION
LINEAR MODEL

>x=c(3.9,2.1,6.4,5.7,4.7,2.8,3.4,7.5,3.0,4.5)
>y=c(6.5,7.8,5.2,8.2,9.2,8.9,7.3,9.8,5.6,4.6)
>a=lm(y~x)
>summary(a)

6
7
OUTPUT
Call:
lm(formula = y ~ x)
Residuals:
Min 1Q Median 3Q Max
-2.7255 -1.3034 0.4168 1.5894 2.0108
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6.6299 1.7085 3.881 0.00467 **
x 0.1546 0.3642 0.424 0.68244
---
8
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.
’ 0.1 ‘ ’ 1
Residual standard error: 1.873 on 8 degrees of free
dom
Multiple R-squared: 0.02202, Adjusted R-squared:
-0.1002
F-statistic: 0.1801 on 1 and 8 DF, p-value: 0.6824

MODEL: Y= 6.6299+0.1546x

9
R-code for Quadratic
>x=c(numeric values)
>y=c(numeric values)
>b=lm(y~x+I(x^2))
>summary(a)

• Here y is the dependent variable

• X is the independent variable

10
QUADRATIC

>x=c(3.9,2.1,6.4,5.7,4.7,2.8,3.4,7.5,3.0,4.5)
>y=c(6.5,7.8,5.2,8.2,9.2,8.9,7.3,9.8,5.6,4.6)
>b=lm(y~x+I(x^2))
>summary(b)

11
12
Call:
lm(formula = y ~ x + I(x^2))
Residuals:
Min 1Q Median 3Q Max
-2.38024 -1.26603 0.08659 1.09552 2.57011
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 11.8590 4.8338 2.453 0.0439
*
x -2.3402 2.1926 -1.067 0.3213
I(x^2) 0.2612 0.2265 1.153 0.2867
13
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05
‘.’ 0.1 ‘ ’ 1

Residual standard error: 1.836 on 7 degrees of

freedom
Multiple R-squared: 0.1781, Adjusted R-squared:
-0.05667
F-statistic: 0.7587 on 2 and 7 DF, p-value: 0.5032

MODEL:𝐲 = 𝟏𝟏. 𝟖𝟓𝟗 − 𝟐. 𝟑𝟒𝟎𝟐𝒙 + 𝟎. 𝟐𝟔𝟏𝟐𝒙𝟐

14
R-Code for Cubic
>x=c(numeric values)
>y=c(numeric values)
>d=lm(y~x+I(x^2)+I(x^3))
>summary(d)

• Here y is the dependent variable

• X is the independent variable

15
CUBIC
>x=c(3.9,2.1,6.4,5.7,4.7,2.8,3.4,7.5,3.0,4.5)
>y=c(6.5,7.8,5.2,8.2,9.2,8.9,7.3,9.8,5.6,4.6)
>c=lm(y~x+I(x^2)+I(x^3))
>summary(c)

16
17
Call:
lm(formula = y ~ x + I(x^2) + I(x^3))
Residuals:
Min 1Q Median 3Q Max
-2.0692 -1.4149 0.1101 1.2082 2.5999
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6.78399 14.68099 0.462 0.660
x 1.44018 10.50169 0.137 0.895
I(x^2) -0.59553 2.33259 -0.255 0.807
I(x^3) 0.05974 0.16178 0.369 0.725
18
Residual standard error: 1.961 on 6 degrees of
freedom
Multiple R-squared: 0.1964, Adjusted R-squared: -
0.2054
F-statistic: 0.4888 on 3 and 6 DF, p-value: 0.7026

MODEL:𝐲 = 6.78399+1.44018x−0. 𝟓𝟗𝟓𝟓𝟑𝒙𝟐 +0.𝟎𝟓𝟗𝟕𝟒𝒙𝟑

19
R-Code for QUARTIC
>x=c(numeric values)
>y=c(numeric values)
>e=lm(y~x+I(x^2)+I(x^3)+I(x^4))
>summary(e)

• Here y is the dependent variable

• X is the independent variable

20
QUARTIC
>x=c(3.9,2.1,6.4,5.7,4.7,2.8,3.4,7.5,3.0,4.5)
>y=c(6.5,7.8,5.2,8.2,9.2,8.9,7.3,9.8,5.6,4.6)
>d=lm(y~x+I(x^2)+I(x^3)+I(x^4))
>summary(d)

21
22
Call:
lm(formula = y ~ x + I(x^2) + I(x^3) + I(x^4))
Residuals:
1 2 3 4 5 6
-0.5671 -0.4279 -1.3427 1.5557 2.0758 1.9301

7 8 9 10
0.3901 0.2268 -1.2881 -2.5526

23
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 36.57466 46.70585 0.783 0.469
x -28.93120 46.28444 -0.625 0.559
I(x^2) 10.22251 16.19846 0.631 0.556
I(x^3) -1.54560 2.38226 -0.649 0.545
I(x^4) 0.08439 0.12491 0.676 0.529

Residual standard error: 2.056 on 5 degrees of freedom

Multiple R-squared: 0.2636,Adjusted R-squared: -
0.3255
F-statistic: 0.4475 on 4 and 5 DF, p-value: 0.772

MODEL:y= 36.5747−28.93120x+10.22𝟐𝟓𝒙𝟐 −1.5𝟒𝟓𝟔𝒙𝟑 +0.𝟎𝟖𝟒𝟑𝟗𝒙𝟒

24
 MULTIPLE REGRESSION
A regression with two or more explanatory variable is called a multiple
regression. Rather than modelling the mean as a straight line as in linear
regression, it is now modelled as a function of several explanatory variables.

25
R-code for Multiple Regression
>explanatory1=c(numeric values)
>explanatory2=c( numeric values)
>response=c(numeric values)
>s=lm(response~explanatory1+explanatory2)
>summay(s)
Here the terms response and explanatory1&2 should be replaced with names
of the response and explanatory variables respectively.
NOTE: The explanatory variables can be more than two.

26
 EXAMPLE
The electric power consumed each month by a chemical plant is thought
to be related to the average ambient temperature(𝑥1 ) and tons of
product produced(𝑥2 ).The past 8 month historical data is presented
below:

𝑥1 91 90 88 87 91 94 87 86

𝑥2 25 21 24 25 25 26 25 25

Y 240 236 270 274 301 316 300 296

Fit a multiple regression to the set of data above.

27
 SOLUTION
> x1=c(91,90,88,87,91,94,87,86)
> x2=c(25,21,24,25,25,26,25,25)
> y=c(240,236,270,274,301,316,300,296)
> a=lm(y~x1+x2)
> summary(a)

28
29
 SOLUTION
Call:
lm(formula = y ~ x1 + x2)

Residuals:
1 2 3 4 5 6
-45.718 4.871 -2.461 -12.285 15.282 17.024
7 8
13.715 9.573

30
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -43.4558 331.1189 -0.131 0.9007
x1 -0.1417 3.4741 -0.041 0.9690
x2 13.6828 6.2329 2.195 0.0796 .
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘
’ 1

Residual standard error: 24.8 on 5 degrees of freedom

Multiple R-squared: 0.4927, Adjusted R-squared: 0.2897
F-statistic: 2.428 on 2 and 5 DF, p-value: 0.1833

MODEL:y=−43.4558−0.𝟏𝟒𝟏𝟕𝒙𝟏 +13.6828𝒙𝟐
31
When there is interaction among the
independent variables.
The R-Code command is:
> x1=c(numeric values)
> x2=c(numeric values)
>y=c(numeric values)
>b=lm(y~x1+x2+x1*x2)
> summary(b)

32
 EXAMPLE
The electric power consumed each month by a chemical plant is thought
to be related to the average ambient temperature(𝑥1 ) and tons of
product produced(𝑥2 ).There is an interaction between ambient
temperature and product produced. The past 8 month historical data is
presented below:

𝑥1 91 90 88 87 91 94 87 86

𝑥2 25 21 24 25 25 26 25 25
Y 240 236 270 274 301 316 300 296

Fit a multiple regression.

33
When there is interaction among the independent variables.

> x1=c(91,90,88,87,91,94,87,86)
> x2=c(25,21,24,25,25,26,25,25)
>y=c(240,236,270,274,301,316,300,296)
>b=lm(y~x1+x2+x1*x2)
> summary(b)

34
35
OUTPUT
Call:
lm(formula = y ~ x1 + x2 + x1 * x2)

Residuals:
1 2 3 4 5 6
-36.003 5.462 -14.565 -10.977 24.997 7.283
7 8
15.023 8.780

36
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 15661.115 19353.552 0.809 0.464
x1 -174.234 214.539 -0.812 0.462
x2 -607.238 765.100 -0.794 0.472
x1:x2 6.880 8.477 0.812 0.463

Residual standard error: 25.69 on 4 degrees of freedom

Multiple R-squared: 0.5644, Adjusted R-squared:
0.2377
F-statistic: 1.727 on 3 and 4 DF, p-value: 0.299

MODEL:y=15661.1−1𝟕𝟒. 𝟐𝟑𝟒𝒙𝟏 −607.2𝒙𝟐 + 6.88𝒙𝟏 𝒙𝟐

37
 EXPONENTIAL REGRESSION
Example
The data shows the cooling temperatures of a freshly brewed cup of coffee
after it is poured from the brewing pot into a serving cup. The brewing pot
temperature is approximately 180º F.

Time 0 1 8 11 15 18 22 25 30 34 38 42 45 50

Temp 179.5 168.7 158.1 149.2 141.7 134.6 125.4 123.5 116.3 113.2 109.1 105.7 102.2 100.5

Find the law in the form T = 𝑎𝑏 𝑡

38
R-Code
>time=c(0,5,8,11,15,18,22,25,30,34,38,42,45,50)
>temp=c(179.5,168.7,158.1,149.2,141.7,134.6,125.4,123.5,116.3,113.2,
109.1,105.7,102.2,100.5)
>a=lm(log(temp)~time)
>summary(a)

39
40
 SOLUTION
Call:
lm(formula = log(temp) ~ time)

Residuals:
Min 1Q Median 3Q Max
-0.052753 -0.025261 -0.005929 0.014306 0.056930

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.1443601 0.0172782 297.74 < 2e-16 ***
time -0.0118227 0.0005988 -19.74 1.62e-10 ***
41
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05
‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.03415 on 12 degrees of

freedom
Multiple R-squared: 0.9701, Adjusted R-squared:
0.9676
F-statistic: 389.8 on 1 and 12 DF, p-value:
1.621e-10

y= 5.144-0.01182log(time)
42
 Example 2
The periods and mean distances of some of the planets are given below:

Period P days 87.97 224.7 365.3 687.0 4333 10760

Mean distance s in 58 108 150 228 778 1426
millions of km

Find a law in the form 𝑃 = 𝑘𝑠 𝑛 .

43
R-Code
>p=c(87.97,224.7,365.3,687.0,4333,10760)
>s=c(58,108,150,228,778,1426)
>a=lm(log(p)~log(s))
>summary(a)

44
45
Call:
lm(formula = log(p) ~ log(s))

Residuals:
1 2 3 4
-9.492e-04 3.871e-03 -3.153e-03 1.154e-04
5 6
-9.961e-05 2.155e-04
46
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.6154480 0.0052541 -307.5 6.71e-10 ***
log(s) 1.5006720 0.0009335 1607.5 8.99e-13 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘
’ 1

Residual standard error: 0.002544 on 4 degrees of freedom

Multiple R-squared: 1, Adjusted R-squared: 1
F-statistic: 2.584e+06 on 1 and 4 DF, p-value: 8.986e-13

𝒑 = −𝟏. 𝟔𝟏𝟓𝟒𝟒𝟖 + 𝟏. 𝟓𝟎𝟎𝟔𝟕𝟐𝒏

47
 Example 3
The width of successive whorls of a shell of Turbo duplicatus have been
measured:

Positions of whols n 1 2 3 4 5 6 7 8
Width of whorl w cm 3.33 2.84 2.39 2.03 1.70 1.45 1.22 1.04

Find the law in the form 𝑤 = 𝑎𝑏 𝑛 .

48
R-Code
>n=c(1,2,3,4,5,6,7,8)
>w=c(3.33,2.84,2.39,2.03,1.70,1.45,1.22,1.04)
>r=lm(log(w)~n)
>summary(r)

49
50
Call:
lm(formula = log(w) ~ n)

Residuals:
Min 1Q Median 3Q
Max
-0.0065511 -0.0033214 0.0006323 0.0036527
0.0049239

51
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 1.3733474 0.0035109 391.2 1.88e-14 ***
n -0.1672336 0.0006953 -240.5 3.48e-13 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘
’ 1

Residual standard error: 0.004506 on 6 degrees of freedom

Multiple R-squared: 0.9999, Adjusted R-squared:
0.9999
F-statistic: 5.786e+04 on 1 and 6 DF, p-value: 3.484e-13
𝒘 = 𝟏. 𝟑𝟕𝟑𝟑𝟒𝟕𝟒 − 𝟎. 𝟏𝟔𝟕𝟐𝟑𝟑𝟔 𝐥𝐨𝐠 𝒃
52