Scribd
Upload
18 views11 pages

Rounding Results in Linear Regression

The document provides data and questions related to linear regression analysis. It includes sample output from Excel for linear regression on different datasets. Questions ask to calculate slope, intercept, fitted values, residuals and other metrics from the linear regression models.

Uploaded by

NURUL SYAHIRAH BINTI ABDUL HALIM
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
18 views11 pages

Rounding Results in Linear Regression

The document provides data and questions related to linear regression analysis. It includes sample output from Excel for linear regression on different datasets. Questions ask to calculate slope, intercept, fitted values, residuals and other metrics from the linear regression models.

Uploaded by

NURUL SYAHIRAH BINTI ABDUL HALIM
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

TUTORIAL 3 KKKQ2023 ENGINEERING STATISTICS ATHIF2023

QUESTION NO. 1

An article in Technometrics by S. C. Narula and J. F. Wellington (“Prediction, Linear Regression, and a


Minimum Sum of Relative Errors,” Vol. 19, 1977) presents data on the selling price (y) and annual
taxes (x) for 24 houses. The taxes include local, school and county taxes. The data are shown in the
following table.

Sale Taxes
25.9 4.9176
29.5 5.0208
27.9 4.5429
25.9 4.5573
29.9 5.0597
29.9 3.891
30.9 5.898
28.9 5.6039
35.9 5.8282
31.5 5.3003
31 6.2712
30.9 5.9592
30 5.05
36.9 8.2464
41.9 6.6969
40.5 7.7841
43.9 9.0384
37.5 5.9894
37.9 7.5422
44.5 8.7951
37.9 6.0831
38.9 8.3607
36.9 8.14
45.8 9.1416

OUTPUT FROM SIMPLE LINEAR REGRESSION USING EXCEL

Coefficient Standar t Stat P-value Lower Upper Lower Upper


s d Error 95% 95% 95.0% 95.0%
Intercep 13.32018 2.57172 5.17947 3.42E- 7.98675 18.6536 7.98675 18.6536
t 2 9 05 5 5
Taxes 3.324371 0.39027 8.51799 2.05E- 2.51498 4.13375 2.51498 4.13375
6 8 08 8 4 8 4

(a) Calculate the least squares estimates of the slope and intercept.

yˆ = ˆ0 + ˆ1 x
= 13.32018 + 3.324371x
ˆ1 = 3.324 (round off to 3 decimal places)
ˆ0 = 13.32 (round off to 2 decimal places)

(b) Find the mean selling price given that the taxes paid are x = 6.6.

yˆ = 13.32018 + 3.324371(6.6)
= 35.2610286
= 35.26 (round off to 2 decimal places)

(c) Calculate the fitted value of y corresponding to x = 8.2464 (observation #14). Find the
corresponding residual.

Fitted value of y corresponding to x = 8.2464


yˆ = 13.32018 + 3.324371(8.2464)
= 40.73427301
= 40.73 (round off to 2 decimal places)
corresponding residual
e = y − yˆ = 36.9 − 40.73 = −3.83 (round off to 2 decimal places)

ALTERNATIVE SOLUTION
S xy
ˆ1 =
S xx
2
 24   24   24 
  i   i 
x y   xi 
S xx =  xi −  i =1 
24
   
24
S xy =  xi yi − i =1 i =1 2

i =1 24 i =1 24
(153.718)(830.7) (153.718) 2
= 5511.925 − = 1042.114 −
24 24
= 191.360725 = 57.56301983

S xy 191.360725
ˆ1 = = = 3.32436911 = 3.324 (round off to 3 d.p.)
S xx 57.56301983
ˆ0 = y − ˆ1 x
 830.7   153.718 
=  − (3.324)  
 24   24 
= 13.322557
= 13.32 (round off to 2 d.p.)
QUESTION 2

Part 1
Calculate the least squares estimate of the slope. (Round your answer to 3 decimal places.)
S xy
ˆ1 =
S xx

(  x )(  y ) S ( x )
2

S xy =  =x −
i
xy −
i i 2
i i xx i
14 14
(43)(572) (43) 2
= 1697.80 − = 157.42 −
14 14
= −59.05714286 = 25.34857143

S xy −59.05714286
ˆ1 = = = −2.329801623 = −2.330 (round off to 3 d.p.)
S xx 25.34857143

Part 2
Calculate the least squares estimate of the intercept. (Round your answer to 3 decimal places.)
ˆ0 = y − ˆ1 x
 572   43 
=  − (−2.330)  
 14   14 
= 48.01296213
= 48.013 (round off to 3 d.p.)

Part 3
Use the equation of the fitted line to predict what permeability would be observed when the
compressive strength is x = 4.3. (Round your answer to 2 decimal places.)
yˆ = ˆ0 + ˆ1 x
= 48.01296213 + ( −2.329801623)(4.3)
= 37.99481515
= 37.99 (round off to 2 d.p.)

Part 4
Give a point estimate of the mean permeability when compressive strength is x = 3.7. (Round your
answer to 2 decimal places.)
yˆ = ˆ0 + ˆ1 x
= 48.01296213 + ( −2.329801623)(3.7)
= 39.39269612
= 39.39 (round off to 2 d.p.)
Part 5
Suppose that the observed value of permeability at x = 3.7 is y = 46.1. Calculate the value of the
corresponding residual. (Round your answer to 2 decimal places.)
ei = yi − yˆi = 46.1 − 39.39 = 6.71

QUESTION 3

(a) Fit the simple linear regression model using the method of least squares. Find the least
squares estimates of the intercept and slope in the simple linear regression model. Find an
estimate of 𝜎2.
OUTPUT FROM SIMPLE LINEAR REGRESSION USING EXCEL

ANOVA
Significance
df SS MS F F
Regression 1 136.6829 136.6829 1.767522 0.241116
Residual 5 386.6513 77.33027
Total 6 523.3343

Coefficient Standar Lower Upper Lower Upper


s d Error t Stat P-value 95% 95% 95.0% 95.0%
- -
Intercep 32.1053 1.73259 26.903 138.155 26.903 138.155
t 55.62561 9 4 0.14371 9 1 9 1
0.02569 - 0.24111 0.03188 0.03188
Temp -0.03416 3 1.32948 6 -0.1002 8 -0.1002 8

least squares estimates of the intercept


ˆ0 = 55.62561 = 55.626 (3 d.p.)
least squares estimates of the slope
ˆ1 = −0.03416 (5 d.p.)
estimate of 𝜎2
SS E 386.6513
2 = = = 77.33026 = 77.33 (2 d.p.)
n−2 7−2
(b) Estimate the mean porosity for a temperature of 1431 oC.
yˆ = ˆ0 + ˆ1 x
= 55.62561 + ( −0.03416)(1431)
= 6.74265
= 6.74 (2 d.p.)
ALTERNATIVE SOLUTION
S xy
ˆ1 =
S xx
2
 7  7   7 
  i   i 
x y   xi 
S xx =  xi −  i =1 
7
  
7
S xy =  xi yi − i =1 i =1 2

i =1 7 i =1 7
(8700)(92.2) (8700) 2
= 110590 − = 10930000 −
7 7
= −4001.428571 = 117142.8571

S xy −4001.428571
ˆ1 = = = −0.034158536 = −0.03416 (round off to 5 d.p.)
S xx 117142.8571
ˆ0 = y − ˆ1 x
 92.2   8700 
=  − (−0.034158536)  
 7   7 
= 55.62560903
= 55.626 (round off to 3 d.p.)

QUESTION 4
_________________________________________________________________________________
S xy
ˆ1 =
S xx

(  x )(  y ) S ( x ) ( y )
2 2

S xy =  =x − S yy =  y −
i i
xy −
i i 2 2
i i xx i i
14 14 14
(43)(572) (43) 2 (572) 2
= 1697.80 − = 157.42 − = 23530 −
14 14 14
= −59.05714286 = 25.34857143 = 159.7142857

−59.05714286
S xy
ˆ1 = = = −2.329801623 = −2.330 (round off to 3 d.p.)
S xx 25.34857143
MS R SS R /1
f0 = =
MS E SS E / (n − 2)
SS = ˆ S = (−2.329801623)(−59.05714286) = 137.5914273 = 137.59 (2 d.p.)
R 1 xy

SST = SS R + SS E
SS E = SST − SS R
= S yy − SS R
= 159.7142857 − 137.5914273
= 22.1228584
MS R SS R /1 137.5914273 /1
f0 = = = = 74.63308302 = 74.63 (2 d.p.)
MS E SS E / (n − 2) 22.1228584 / (14 − 2)

8) Because 74.63 > 4.75 reject 𝐻0 and conclude that compressive strength is significant in predicting
intrinsic permeability of concrete at 𝛼=0.05. We can therefore conclude that the model specifies a
useful linear relationship between these two variables.

(a) Test for significance of regression using 𝛼=0.05.


Reject the null hypothesis
(b) Estimate 𝜎2.
SS E 22.1228584
ˆ 2 = MS E = = = 1.843571533 = 1.844 (3 d.p.)
n−2 14 − 2

(d) In this model, what is the standard error of the intercept?

ˆ 2 1 x2  1 (43 /14) 2 
se(  0 ) =   +
ˆ  = (1.843571533)  +  = 0.904313862 = 0.904 (3 d.p.)
 n S xx  14 25.34857143 
What is the standard error of the slope?
ˆ 2 1.843571533
se( ˆ1 ) = = = 0.269682802 = 0.270 (3 d.p.)
S xx 25.34857143

QUESTION 5
OUTPUT FROM SIMPLE LINEAR REGRESSION USING EXCEL
ANOVA
Significance
df SS MS F F
Regression 1 92.93353 92.93353 53.5015 8.57E-07
Residual 18 31.26647 1.737026
Total 19 124.2

Standard Upper Lower Upper


Coefficients Error t Stat P-value Lower 95% 95% 95.0% 95.0%
7.83E-
Intercept -10.1315 1.9949 -5.07872 05 -14.3227 -5.94041 -14.3227 -5.94041
8.57E-
x 0.174294 0.023829 7.314472 07 0.124232 0.224356 0.124232 0.224356

MS R 92.93353
f0 = = = 53.50151926 = 53.50 (2 d.p.)
MS E 1.737026

OR ALTERNATIVELY, (CALCULATE BY FORMULA)

2 2
 20  20   20   20 
  i   i 
x y  i  x   yi 
  S yy =  yi −  i =1 
20 20
  
20
S xy =  xi yi − i =1 i =1
S xx =  xi −2 i =1 2

i =1 20 i =1 20 i =1 20
2 2
(1656)(86) (1656) (86)
= 7654 − = 140176 − = 494 −
20 20 20
= 533.2 = 3059.2 = 124.2
S xy 533.2
ˆ1 = = = 0.174293933
S xx 3059.2
86  1656 
ˆ0 = y − ˆ1 x = − (0.174293933)   = −10.13153765
20  20 
SS = ˆ S = (0.174293933)(533.2) = 92.9335251
R 1 xy

SST = SS R + SS E
SS E = SST − SS R
= S yy − SS R
= 124.2 − 92.9335251
= 31.2664749
MS R SS R /1 92.9335251/1
f0 = = = = 53.50150464 = 53.50 (2 d.p.)
MS E SS E / (n − 2) 31.2664749 / (20 − 2)

SS E 31.2664749
ˆ 2 = MS E = = = 1.737026383
n−2 20 − 2

(a) Test for significance of regression using 𝛼=0.05. What is the P-value for this test? Round to 7
decimal places.
P-value = 8.57E-07 or 0.000000857 = 0.0000009 (7 d.p.)

(b) Estimate the standard errors of the slope and intercept.

Standard error of the slope

ˆ ˆ 2 1.737026383
se( 1 ) = = = 0.02382864 = 0.0238 (4 d.p.)
S xx 3059.2

Standard error of the intercept

1 x2   1 (1656 / 20) 2 
se( ˆ0 ) = ˆ 2  +  = (1.737026383)  +  = 1.994899887 = 1.995 (3 d.p.)
 n S xx   20 3059.2 

(c) Test 𝐻0: 𝛽0=0 versus 𝐻1: 𝛽0≠0 using 𝛼=0.05. Find the P-value for this test.

H 0 : 0 = 0
H1 : 1  0

 = 0.05

ˆ0 − ˆ0,0 ˆ0 − ˆ0,0 −10.13153765


t0 = = = = −5.078719849 = −5.079
1 x2  se( ˆ0 ) 1.994899887
ˆ 2  +
n S xx 

Reject H0 if | t0 | > t /2, n − 2


Since | t0 | = 5.079 > t /2, n −2 = t0.025,18 = 2.101

Therefore, we reject H0.

P-value = 7.83E-05 = 0.0000783 = 0.000078 (round off to 6 d.p.)

QUESTION 6

OUTPUT FROM SIMPLE LINEAR REGRESSION USING EXCEL

ANOVA
Significance
df SS MS F F
Regression 1 45.15415 45.15415 40.38286 0.000383
Residual 7 7.827059 1.118151
Total 8 52.98121

Standard Upper Lower Upper


Coefficients Error t Stat P-value Lower 95% 95% 95.0% 95.0%
Intercept 32.04867 2.885174 11.10806 1.07E-05 25.22632 38.87102 25.22632 38.87102
Stress (x) -0.27712 0.043608 -6.35475 0.000383 -0.38023 -0.174 -0.38023 -0.174

P-value = 0.00383 < α=0.01


Thus, we reject H0.
There is evidence of a linear relationship between these two variables.
(b) Estimate the standard errors of the slope and intercept.
2
 9 
  xi 
S xx =  xi −  i =1 
9
2

i =1 9
(591) 2
= 39397 −
9
= 588

ˆ 2 = MS E = 1.11815130026455
ˆ 2 1.11815130026455
se( ˆ1 ) = = = 0.043607543 = 0.044 (3 d.p.)
S xx 588
1 x2   1 (591/ 9) 2 
se( ˆ0 ) = ˆ 2  +  = 1.11815130026455  9 + 588  = 2.885173569 = 2.885
 n S xx   

OR ALTERNATIVELY, (CALCULATE BY FORMULA)

2 2
 9  9   9   9 
 i  i 
x y  i  x   yi 
  S yy =  yi −  i =1 
9 9
   
9
S xy =  xi yi − i =1 i =1
S xx =  xi −2 i =1 2

i =1 9 i =1 9 i =1 9
(591)(124.663) (591) 2 (124.663) 2
= 8023.26 − = 39397 − = 1779.744 −
9 9 9
= −162.9436667 = 588 = 52.98138122
S −162.9436667
ˆ1 = xy = = −0.277115079
S xx 588
124.663  591 
ˆ0 = y − ˆ1 x = − (−0.277115079)   = 32.04866803
9  9 
SS = ˆ S = (−0.277115079)(−162.9436667) = 45.15414715
R 1 xy

SST = SS R + SS E
SS E = SST − SS R
= S yy − SS R
= 52.98138122 − 45.15414715
= 7.827234047

SS E 7.827234047
ˆ 2 = MS E = = = 1.118176292
n−2 9−2
ˆ 2 1.118176292
se( ˆ1 ) = = = 0.04360803 = 0.044 (3 d.p.)
S xx 588
1 x2   1 (591/ 9) 2 
se( ˆ0 ) = ˆ 2  +  = 1.118176292  9 + 588  = 2.885205813 = 2.885
 n S xx   

Common questions

Powered by AI

High residual variance may arise from omitted variables, model misspecification, measurement errors, or high data variability. It implies less precise predictions, indicates model inadequacy, and necessitates identifying unaccounted sources of variation or adjusting the model to better capture data dynamics, impacting the model’s reliability and validity of inferences drawn .

The standard error in linear regression quantifies the variability in the estimate of the regression coefficients, providing a measure of precision. It is determined by taking the square root of the estimated variance of the coefficient, \( \text{SE}(\beta) = \sqrt{\frac{\sigma^2}{S_{xx}}} \) for the slope and \( \sqrt{\sigma^2 \left( \frac{1}{n} + \frac{\bar{x}^2}{S_{xx}} \right)} \) for the intercept .

A predictor’s significance is determined through hypothesis testing of its coefficient (\( \beta \)) using t-tests where the null hypothesis states \( \beta = 0 \). By comparing the calculated t-statistic to critical values or evaluating the P-value against a significance level (\( \alpha \)), you ascertain if the predictor meaningfully enhances the model. A low P-value (\(< \alpha \)) leads to rejecting the null hypothesis, indicating the predictor's significant contribution .

Checking for linear relationships is crucial since linear regression assumes a linear relationship between predictors and response, impacting predictions and model validity. It is verified by inspecting scatterplots for linearity, performing residual plots to detect non-linearity, and using correlation coefficients to measure the strength of linear association .

Estimating the mean response provides the average response value for a given level of the predictor variable within the population, involving less variability and narrower confidence intervals. In contrast, predicting an individual response includes variability around the individual outcome, producing wider prediction intervals due to additional elements of model error and observation error .

The P-value indicates the probability that the observed data would occur by random chance if the null hypothesis were true. In regression analysis, a low P-value (typically < 0.05 or < 0.01 depending on the chosen significance level, \( \alpha \)) suggests that the null hypothesis (i.e., the coefficient being equal to zero) can be rejected, implying the coefficient is statistically significant. For example, a P-value of 0.000383 < \( \alpha = 0.01 \) leads to rejecting the null hypothesis, indicating that the stress variable significantly predicts the outcome .

The F-test in regression compares the model with no predictors (constant model) against the model specified. The calculated F-value is compared to a critical F-value from F-distribution tables at a given significance level. If the calculated F-value is greater than the critical F-value, the null hypothesis that no linear relationship exists is rejected. For example, in a test with \( F = 53.50 \) where the critical \( F_{0.05} = 4.75 \), the null hypothesis is rejected, indicating a significant linear regression relationship .

The residual variance, \( \sigma^2 \), in a regression model is estimated by dividing the residual sum of squares (SS_E) by the degrees of freedom (\( n - 2 \) for simple linear regression). For instance, \( \sigma^2 = \frac{386.6513}{7 - 2} = 77.33026 \).

The least squares estimates of the slope and intercept in a simple linear regression model are calculated using the formulas: for slope \( \beta_1 = \frac{S_{xy}}{S_{xx}} \) and for intercept \( \beta_0 = \bar{y} - \beta_1 \bar{x} \). \( S_{xy} \) and \( S_{xx} \) are the sums of squares of the statistical variables involved, calculated as \( S_{xy} = \sum xy - \frac{(\sum x)(\sum y)}{n} \) and \( S_{xx} = \sum (x^2) - \frac{(\sum x)^2}{n} \) respectively .

Residual analysis helps check for randomness in residuals, which indicates a good model fit. Non-random patterns suggest issues such as non-linearity, heteroscedasticity, or missed variables. Assessment involves plotting residuals vs. fitted values to detect patterns, and standardized or studentized residuals help identify outliers or high-leverage points that could affect the model .

You might also like