0

2
1
3
5
4
 MATEMATIKA STATISTIKA
“Simple Linear Regression (SLR) Model and
Multiple Linear Regression (MLP) Model”
 

Name:Joko Hariaji Name: Novita Sari Sitanggang Name:Putri Anggraini Purba

NIM: 8206172024 NIM: 8206172030 NIM: 8206172020

Lecturers : Dr. Elmanani Simamora, M. Si

PRODI MAGISTER PENDIDIKAN MATEMATIKA

PROGRAM PASCASARJANA UNIMED
DESEMBER, 2020
 A. Background

Changes in the value of a variable do not always happen on their own, but changes in the value of the variable can also be caused by
changes in other variables related to these variables. To find out the pattern of changes in the value of a variable caused by other
variables, an analysis tool is needed that allows us to estimate the value of this variable at a certain value of the variables that
influence it. In statistics, a technique commonly used to analyze the relationship between two or more variables is regression analysis.

Regression analysis is a technique for building straight line equations and using these equations to make estimates. Regression
analysis is a statistical analysis tool that utilizes the relationship between two or more variables. The aim is to make a reliable estimate
for the value of a dependent variable, if the value of the other variable associated with it is known (independent variable).

B. Formulation of the problem C. Objective

1. What is a linear regression? 1. To know the meaning of the What is a linear regression
2. What is Simple Linear Regression (SLR) Model? 2. To know What is Simple Linear Regression (SLR) Model
3. What is Multiple linear regression (MLP) 3. To know What is Multiple linear regression (MLP) Model
Model?
 A. Definition Analysis Regression Linear
 Analysis regression is a a model that attempted explain relationship functional
Among two variable, where one variable Act as predictor to variable response
 Method correlation discuss closeness relationship between two variable. While
method Regression discuss influence from independent variable to variable
dependent.
 Destination analysis regression in a manner general there is three, that is to
measure power relationship, test influence and forecasting/predictions
 For example, we want to discuss how the cost of promotion affects sales. Here
we know that there are two variables. They are sales and promotional
expenses. Well, the dependent variable is sales, while the independent variable
is promotion costs.
Terms Variable in regression linear
• Variables X and Y have scale measurement o at least interval
• Data collected must Fulfill model basic
• Collection the data can done through two way:
• Variable X determined more so first
• All at once counted X and Y
• Because price variable X fixed, then distribution variable Y same with
variable
 Model Regression Linear
• In
  general, regression is divided into two, simple regression and multiple
regression. Called Simple Regression if there is only one independent variable,
while Multiple Regression if there is more than one independent variable.
• Basic Model Regression Linear Simple
Y=
Basic Model Regression Linear Multiple
Y=
Where
Y is variable dependent / response / it depends x, , , , , is variable independent/
predictor /free
intercept parameter (constant)

is the residual is a random variable

 B. Regression Iinear Simple
•  Generally, equation basic regression linear simple can be written with:
= a + bx
Following is Steps testing analysis regression linear simple:
1. Assessment score parameters a and b could use method square the
smallest (Ordinary Least Square)
a=
b=
 Step Testing Regression Linear simple
•2.  Testing Partial Significance Result Assessment parameter
Use it for test significance parameter what score constant (a) and
Coefficient (b) take effect to variable response / dependent.
Previous need counted score variance from x, y and error.
= =

= (-.)
 Step Testing Regression linear simple
•  Test Significance Constant a
A.
 Hypothesis :
= = 0 constantsta not take effect significant
= 0 constantsta take effect significant
 Specify level significance {looking for table t with df = n-2}
 Statistics Test:
t = = it depends on hypothesis early or assumption zero (0)
 Criteria Test: If , then rejected
If , then be accepted
 Conclusion
 Step Testing Regression linear simple
•  Test Significance Constant b
B.
 Hypothesis :
= = 0 coefficient not take effect significant
= 0 coefficient take effect significant
 Specify level significance
{looking for table t with df = n-2}
 Statistics Test:
t = = it depends on hypothesis early or assumption zero (0)
 Criteria Test: If , then rejected
If , then be accepted
 Conclusion
 Step Testing Regression Linear simple
•3.  Testing simultaneously Significance Result Estimator Parameter
 Hypothesis :
: variable independent not take effect to variable dependent
: variable independent take effect to variable dependent
 Specify level significance {looking for table F with d= 1, d= n-2}
 Statistics Test:
amount Variation amount Square Degree Freedom average amount F count
Square
Regression JKR 1 RJKR F count

Error JKG n-2 RJKG

Total JKT N-1

JKT = amount Square Total = RJKG = Average amount Square Error =

JKR = amount Square Regression= b [- ] =
 Test Criteria:
JKG = amount Square Error = JKT - JKR  If , then rejected
If , then be accepted
RJKR = Average average amount Caudrat Regression =
 Step-step Testing Regression Linear Simple
•4.  Coefficient Correlation and Coefficient Determination
Coefficient Correlation used to measure level closeness relationship among
independent variable to variable dependent
Coefficient determination used to measure how much big influence that given
independent variable to variable dependent.
The measurement could used formula correlation pearson or with formula
below this: Coefficient
Determination
= b.
Coefficient
Correlation
Example Case
Want known influence score exam to score end, to that taken sample as much 12 students SMPN 1 Medan class IX,
noted score exam and score end each, following is the data:
No. Score exam (x) Final Grade (y) Count:
1. 65 85 1. Write it down equation the regression, then Explain mean from equation the!
2. Draw it linearity from data, then conclude!
2. 50 74 3. Test significance Partial score constant and coefficient the regression? Use level real 0.05
3. 55 76 (5%).
4. Test significance simultaneously the result assessment parameter? Use level real 0.05 (5%)
4. 65 90 5. How many I can relationship that occur between score exam with score end?
5. 55 85
6. How much score coefficient determination? Explain mean from score the!
6. 70 87

7. 65 94

8. 70 98

9. 55 81

10. 70 91

11. 50 76

12. 55 74
Answer
1. Equation The regression and Interpretation

No. Score Exam (x) Score end (y) x*y

1. 65 85 4,225 7,225 5,525

2. 50 74 2,500 5,476 3,700

3. 55 76 3,025 5,776 4,180

4. 65 90 4,225 8,100 5,850 = 725

= 1011
5. 55 85 3,025 7,225 4,675
= 44,475
6. 70 87 4,900 7,569 6,090 = 85,905
7. 65 94 4,225 8,836 6.110 = 61,685
n = 12
8. 70 98 4,900 9,604 6,860 = = = 60.42
9. 55 81 3,025 6,561 4,455
= = = 84.25
10. 70 91 4,900 8,281 6,370

11. 50 76 2,500 5,776 3,800

12. 55 74 3,025 5,476 4,070

amoun 725 1,011 44,475 85,905 61,685

t
Answer = 725, = 1.011, = 44,475,
= 85,905, = 61,685, n = 12
= 60.42, = 84.25
•   Equation The regression and Interpretation
1.
Y = a + bx
b = = = = 0.897

a = = (84.25) - (0.897). (60.42) = (84.25) - (54.197) = 30.053

Then equation regression linear ones obtained is;

Y = a + bx = 30.053 + 0.897x
From equation above, could interpreted:
 constant a = 30,053Declare average score end students moment exam (X) zero is
amounting to 30,053
 coefficient b = 0.897Every increase amounting to 1 on score exam (X), then score end
(Y) college student possibility will increase amounting to 30,053
Answer
2. Diagram Scatter (Scatter Plot) Linearity data
120

"Relationship between score exam with

100 score end is proportional straight" 98
94
90 91

85 85
() 87
84.25
81
80 76 76
74 74
Nilai Akhir

60

40

20

30,053 (Konstanta)
0
45 50 55 60 65 70 75

Nilai Ujian
3. Test Significance Partial Parameter constant a and coefficient b
 = 725
= 1011
= 44,475
= 85,905, = 61,685, n = 12

=• = = = = 61,174
= = 7,821
= = = = = 66,205
= = 8,137
= (-.) = . [66,205- (). (61,174)] =[66,205- (0,805) (61,174)]
= (1,1). (66,205-49,245) = (11). (16.96) = 18.66
= = 4,319
a). Test significance constant a
•1). Hypothesis:
= = 0 constantsta not take effect significant to score end
= 0 constant effect significant to the final score
a = 30,053
2). Level Significance = 7,821
= 4,319
= 5% = 0.05 / 2 = 0.025see Table t = 44,475

df = n-2 = 12-2 = 10 = 2,228

3). Statistics Test
t= = = =
= = 2,965
4). Criteria Test
If , then rejected
5). Conclusion:
If , then be accepted Because 2,965> 2,228 then Ho rejected
"" Constant take effect significant to score end
a). Test significance coefficient b b = 0.897
•1).  Hypothesis: = 7,821
= 4,319
= = 0 coefficient score exam not take effect significant to score end
= 0 coefficient score exam significant effect on the final score
2). Level Significance
= 5% = 0.05 / 2 = 0.025see Table t
df = n-2 = 12-2 = 10 = 2,228
3). Statistics Test
t= = = =
t = 5,388
4). Criteria Test
If , then rejected
5). Conclusion:
If , then be accepted Because 5,388 > 2,228 then Ho rejected
"" Coefficient score exam take effect significant to score end
4. Test significance Silmultan the result Assessment parameter = 1011
•   Hypothesis:
1). = 85,905
= independent not take effect to variable dependent
= independent take effect to variable dependent
2). Level significance
= 5% = 0.05 see table F
= 1; = 12-2 = 10 = 4.96
3). Statistics Test
amount amount Square Level Freedom Average amount Square F Count
Variation
Regression JKR 1 RJKR F count
Error JKG n-2 RJKG
Total JKT n-1

JKT = = (85,905) = (85,905) = 85,905-85,176.75 = 728.25

JKR = b [- ] = (0.897) ((61685) - ] =[(0.897) [(61,685) - ]
4. Test significance Silmultan the result Assessment parameter = 1011
•JKR =(0.897) [(61,685) - (61,081.25)]= (0.897). (603.75) = 541.564 = 85,905

JKG = JKT-JKR = 728.25 - 541,564 = 186,686

RJKR = amount amount Level Average amount F Count
RJKG = = = 18,669 Variation Square Freedom Square
Regression 541,564 1 641,564 29,009
4). Criteria Test Error 186,686 10 18,669
• If , then rejected Total 728.25 11

• If , then be accepted
5). Conclusion
Because 29.009> 4.96 then rejected
" Variable independent take effect to variable dependent
Answer
5. Coefficient Correlation among Score Exam (X) with Final Grade (Y).

•To measure level closeness relationship, could use formula correlation pearson or
with formula below this:
= b. = 0.862
Based on criteria Guilford, relationship that occur Among score exam and score
end is "strong"
Answer:
6. Coefficient Determination()
•
Coefficient determination is a square from correlation

= = 0.743 = 74.3%
Score exam take effect to score end amounting to 74.3%. The rest (100% -74.3% = 25.7%) score end
influenced by another factor not described in model
7. Predictions
How many estimate score end a students if score the test is 75?
X = 75
Y = a +bx
Y = 30.053 + 0.897x
Y = 30.053 + 0.897 (75)
Y = 30,053 + 67,275
Y = 97.328
If students have score exam amounting to 75, then estimate score finally is amounting to 97,328
 D. Multiple Linear Regression Analysis and Classic Assumption Test with SPSS

1. Multiple linear regression analysis aims to determine the effect of two or more
independent variables (X) on the dependent variable (Y).

2. According to V. Wiratna Sujarweni (2014; 181) The multiple linear regression model
can be called a good model (has accuracy in estimation, is unbiased and consistent) if
the model meets the assumptions of normality and is free from the classic assumptions
of multicollinearity. , heteroscedasticity, and autocorrelation (time series data)

For more details, tips go to examples of discussion problems

 ROA Stock
ROE (X2) PER (X3)
Banking (X1) Price (Y)
SAMPLE CASE No. Year
Sector
(%) (%) (Rp) (Rp)
1. Research Title: 1 BCA 2011 2.83 25.74 18.02 8000
2012 2.65 22.58 18.99 9100
• Effect of Return On Assets (X1), Return On Equity (X2) 2013 2.87 22.29 16.61 9600
2014 2.99 21.19 19.16 13125
and Price Earning Ratio (X3) on Stock Prices (Y) in 2 BNI 2011 1.94 15.35 12.22 3800
2012 2.11 16.19 9.79 3700
Banking Companies Listed on the Indonesia Stock 2013 2.34 19.00 8.13 3950
2014 2.60 17.75 10.50 6100
3 BRI
Exchange Period 2011 - 2014. Time series secondary data. 2011
2012
3.21
3.39
30.28
28.80
11.03
9,17
6750
6950
2013 3.41 26.92 8.37 7250
Secondary time series data
2. Analysis Stages 2014 3.02 24.82 11.85 11650
4 DANAMO 2011 2.52 16.17 11.55 4100
N 2012 2.43 13.35 13.14 5650
a) Prepare the Research Data Tabulation. 2013 2.64 14.33 8.70 3775
2014 2.26 13.18 16.16 4525
5 INDEPEND 2011 2.30 20.26 12.81 6750
b) Multiple Linear Regression Analysis+Classical Assumption ENT 2012 2.52 20.96 11.77 8100
2013 2.57 21.21 9.37 7850
Test (Normality, Multicollinearity, Heteroscedasticity, and 2014 2.42 19.70 12.17 10775
6 CIMB 2011 1.90 17.29 9.68 1220
Autocorrelation) with SPSS. 2012 2.15 18.76 6.51 1100
2013 1.96 16.60 5.17 920
2014 1.01 8.24 8.98 835
c) Seeing the basis for decision making in the normality test, 7 MAYBANK 2011 0.71 8.44 14.17 420
2012 1.05 12.53 13.16 405
multicollinearity, heteroscedasticity, autocorrelation and 2013 1.12 12.66 11.78 310
2014 0.50 4.86 12.32 208
multiple linear regression (t test and F test). 8 JEWEL 2011 1.36 12.68 9,18 1326
2012 1.14 12.66 6.70 1317
2013 1.04 10.95 6.47 1260
d) Discussion and Conclusion. 2014 1.04 12.22 10.89 1505
3. Basis for Decision Making on Probability Plot Normality Test
According to Imam Ghozali (2011: 161) The regression model is said to be normally distributed if the plotting data (points) that describe the actual data follows a diagonal line.

Conclusion Normality Test

 The regression model is normally distributed based on the image

 Alternative with the smirnov komogorov test
4. The basis of multicollinearity test decision making with the Tolerance and Vif method
According to Imam Ghozali (2011: 107-108) There are no symptoms of multicollinearity, if the Tolerance value is> 0.100 and the VIF
value is <10.00. Coefficientsa
Standardiz
ed
Unstandardized Coefficient
Coefficients s Collinearity Statistics
Model B Std. Error Beta t Sig. Tolerance VIF
1 (Constant) -7026,584 1350,701   -5,202 , 000    
ROA (X1) 2812,565 920,905 , 614 3,054 , 005 , 191 5,246
ROE (X2) 77,417 121,984 , 126 , 635 , 531 , 195 5,141
PER (X3) 391,547 94,634 , 373 4,137 , 000 , 949 1,053
a. Dependent Variable: Stock Price (Y)

Conclusion Multicollinearity Test

Based on the coefficients table, the Tolerance ROA, ROE and PER values ​are more than 0,100 and the VIF values ​both
ROA, ROE and PER are below 10,00 so it can be concluded that there are no symptoms of multicollinearity.
5. Basis for Decision Making on Scatterplots Heteroscedasticity Test
According to Imam Ghozali (2011: 139) There is no heteroscedasticity, if there is no clear pattern (wavy, widened then narrowed) in the
scatterplots image, and the dots spread above and below the number 0 on the Y axis.
Conclusion Heteroscedasticity Test

• Based on the scatterplot image, it appears that the data is

spread below and above zero, so it can be concluded that there
are no symptoms of heteroscedasticity.

• Alternatives with the Glejser test (please try)

6. Durbin Watson's Autocorrelation Test Decision Making Basis
Model Summary b
According to Imam Ghozali (2011: 111) There is no autocorrelation symptom, if the DurbinAdjusted
Watson value
R lies between
Std. du to (4-du).
Error of
Model R R Square Square the Estimate Durbin-Watson
1 , 886a , 784 , 761 1827,368 1,671
a. Predictors: (Constant), PER (X3), ROE (X2), ROA (X1)
b. Dependent Variable: Stock Price (Y)

Discussion and Conclusion Autocorrelation Test

Based on the table and figure of the Durbin Watson
distribution table, it can be concluded that
The value of du is sought in the Watson durbin table value
distribution based on k the number of independent
variables (3) and N a lot of data (32) with a significance of
5%. du (1,650) <Durbin Watson (1,671) <4-du (2,350) so
there is no autocorrelation symptom.
7. Basis of Decision Making on Partial T Test (Multiple Linear Regression) Based on Significance Value
According to Imam Ghozali (2011: 101) if the value is Sig. <0.05, it means that the independent variable
(X) partially affects the dependent variable (Y).
Coefficientsa

Standardize
Unstandardized d
Coefficients Coefficients Collinearity Statistics
Model B Std. Error Beta t Sig. Tolerance VIF
1 (Constant) -7026,584 1350,701   -5,202 , 000    
ROA (X1) 2812,565 920,905 , 614 3,054 , 005 , 191 5,246
ROE (X2) 77,417 121,984 , 126 , 635 , 531 , 195 5,141
PER (X3) 391,547 94,634 , 373 4,137 , 000 , 949 1,053
a. Dependent Variable: Stock Price (Y)

Conclusion of the Partial T Test

 Return On Assets (X1) affects the Stock Price (Y) because of the sig. ROA <0.05
 Return On Equity (X2) has no effect on Stock Price (Y) because the sig.ROE value is> 0.05
 Price Earning Ratio (X3) affects the stock price (Y) because the sig.PER value <0.05
Decision Making Basis for Partial T Test (Multiple Linear Regression) Based on Calculated Values ​and Tables
According to V. Wiratna Sujarweni (2014; 155), if the value of t count> t table, it means that the independent variable (X) partially
affects the dependent variable.
Coefficientsa
Standar
dized
Unstandardized Coefficie Collinearity
Coefficients nts Statistics
Std. Toleranc
Model B Error Beta t Sig. e VIF
1 (Constan - 1350,70
t) 7026,58 -5,202 , 000  
1   
4
ROA (X1) 2812,56
920,905 , 614 3,054 , 005 , 191 5,246
5
ROE (X2) 77,417 121,984 , 126 , 635 , 531 , 195 5,141
PER (X3) 391,547 94,634 , 373 4,137 , 000 , 949 1,053
a. Dependent Variable: Stock Price (Y)
Based on the table, the t value is ROA = 3.054, ROE = 0.635, PER = 4.137
(Y). The formula for finding T-table = (a / 2; nk-1) = (0.05 / 2; 32-3-1) = (0.025; 28) = 2.048 or if
using the following table;

 Return On Assets (X1) has an effect on stock prices (Y) because the value of T count ROA
(3.054)> T-table (2.048)
 Return On Equity (X2) has no effect on stock prices (Y) because the value of T count ROE
(0.635) <T-table (2.048)
 Price Earning Ratio (X3) affects the stock price (Y) because the value of T count PER (4.137)
<T-table (2.048)
Seeing the Comparison of T Value with Curve

Partial T Test Conclusion:

X1 has a positive effect on Y
X2 has no effect on Y
X3 has a positive effect on Y
 8. Basis of Simultaneous F Test Decision Making (Multiple Linear Regression) Based on Significance
ANOVAa
Value
According to Imam Ghozali (2011: 101) if the value is Sig. <0.05, Sum of Mean
Model Squares df Square F Sig.
it means that the independent variable (X) simultaneously 1 Regression 3401496 1133832
3 33,954 , 000b
88,244 29,415
affects the dependent variable (Y). Residual 9349969 3339274,
28    
4,631 808
Total 4336493
31      
82,875
a. Dependent Variable: Stock Price (Y)
b. Predictors: (Constant), PER (X3), ROE (X2), ROA (X1)

Conclusion of Simultaneous F Test

Based on the ANOVA table, it is found that the sig. simultaneous independent variable is 0.00 <0.05, so it can be concluded
that; Return On Assets (X1), Return On Equity (X2) and Price Earning Ratio (X3) simultaneously affect the stock price (Y).
Basis for Simultaneous F Test Decision Making (Multiple Linear Regression) Based on Calculated Values ​and Tables
According to V. Wiratna Sujarweni (2014; 154), if the value of Fcount> Ftable, it means that the independent variable (X)
simultaneously affects the dependent variable (Y). The formula for finding t-table = (k; nk) = (3; 32-3) = (3; 29) = 2.92
Conclusion of Simultaneous F Test
Because the value of f-count (33,954)> f table (2.92) then, Return On
Assets (X1), Return On Equity (X2) and Price Earning Ratio (X3)
simultaneously affect the Stock Price (Y)
A. Conclusion
CHAPTER III
CLOSING

Linear regression is divided into two, namely simple linear regression and multiple linear regression, simple linear
regression aims to determine the effect of one independent variable on the dependent variable while multiple linear
regression aims to determine the effect of two or more independent variables on the dependent variable.

The requirements for data can be tested using simple linear regression, which must meet the classical assumptions, namely;
valid, reliable, normal and linear. There are two ways of making decisions in the simple regression test; (1) comparing the
significant value with a probability value of 0.005 (5%), (2) comparing the t value with the t table.

Whereas in multiple linear regression the data can be tested if it meets the classical assumptions, namely; normality,
multicollinearity, heteroscedasticity and autocorrelation with sps. the basis of decision making in multiple linear regression
in tests of normality, multicolearity, heteroscedasiticity, autocorrelation and multiple linear regression (t test and F test)

B. Suggestions

1. The reader needs skills in the use of SPSS media / applications because it is very helpful in calculating large amounts of
data.
2. It is necessary to use the R studio application in data calculation, but because the author has not mastered the R studio
application, the focus is more on deepening SPSS.
3. Authors and readers should be able to read T-tables, F-tables and Durbin Watson's value distribution tables, thus helping in
data calculations
TERIMA KASIH