Applying Regression Using e Views (With Commands)

1|E C O N O M E T R I C S 2 T E R M P A P E R
APPLIED ECONOMICS RESEARCH CENTRE
ECONOMETRICS 2
TERM PAPER
SAFIA ASLAM
2013-14
CONTENTS
Chapter 8 Multiple Regression Analysis: The Problem of Inference R e s t r i c t e d Least Squares: Testing Linear Equality Restrictions W a l d Test T e s t i n g For Structural or Parameter Stability of Regression Models O The Chow Test C o m p u t e r Application using EViews Chapter 10 Multicollinearity: What Happens If The Regressors Are Correlated? The Nature of Multicollinearity Sources of Multicollinearity Practical Consequences of Multicollinearity Detection of Multicollinearity O High R2 but Few Significant T-Ratios O High Pair-wise Correlations among Regressors O Examination of Partial Correlations O Auxiliary Regressions O Tolerance and Variance Inflation Factor O Remedial Measures O Do Nothing Or O Follow Some Rules Of Thumb. O A Priori Information O Combining Cross-Sectional and Time Series Data O Dropping a Variable(S) and Specification Bias O Transformation of Variables O First Difference Form O Ratio Transformation
A d d i t i on a l or New Data O Computer Application using EViews Chapter 12 Autocorrelation: What Happens If the Error Terms Are Correlated? The Nature of Autocorrelation Sources of Autocorrelation O Inertia. O Specification Bias: Excluded Variables Case O Specification Bias: Incorrect Functional Form O Cobweb Phenomenon O Lags. O Manipulation Of Data O Data Transformation Practical Consequences of Autocorrelation
Detection of Autocorrelation G r a p h i c a l Method T h e Runs Test D u r b i n Watson D Test A General Test of Autocorrelation: The BreuschGodfrey (Bg) Test R e m e d i a l Measures of Autocorrelation N e w e y West Method. G e n e r a l i z e d Least-Square (G|LS) Method. Correcting For (Pure) Autocorrelation: The Method of Generalized Least Squares (GLS) O When Is Known: O When Is Not Known: The First-Difference Method: O Berenblutt Webb Test, Based On DurbinWatson D Statistic: Estimated From The Residuals: Theil-Nagar Estimate Based On D Statistic: Estimating : The CochraneOrcutt (CO) Iterative Procedure: Estimating : Durbins Two-Step Method: The Durbin h Statistic Computer Application using EViews Chapter 11 Heteroscedasticity: What Happens If the Error Variance Is Non constant? The Nature of Heteroscedasticity Sources of Heteroscedasticity Error-Learning Models Data Collecting Techniques Improve Outliers.
2 i
Is Likely To Decrease.
Other Sources of Heteroscedasticity: O Incorrect Data Transformation (E.G., Ratio or First Difference Transformations) O Incorrect Functional Form (E.G., Linear Versus LogLinear Models). Practical Consequences of Heteroscedasticity Detection of Heteroscedasticity Informal Methods
Nature Of The Problem Graphical Method Formal Methods O Park Test O Glejser Test O Goldfeld-Quandt Test W h i t e s General Heteroscedasticity Test K o e n k e r Bassett (Kb) Test Remedial Measures 1. When i2 is known The method of weighted least square 2. When i2 is not known Whites heteroscedasticity-Consistent Variances and Standard Errors Plausible Assumptions about Heteroscedasticity Pattern.
Assumption 1: The Error Variance Is Proportional To Xi2 Assumption 2: The Error Variance Is Proportional To Xi. The Square Root i Transformation Assumption 3: The Error Variance Is Proportional To the Square Of The Mean Value Of Y. Assumption 4: A Log Transformation 3.
Computer Application using EViews
Chapters 18 To 20 Simultaneous Regression Models
1. The Nature Of Simultaneous-Equation Models 2. The Identification Problem 3. Rules For Identification The Order Condition Of Identifiability The Rank Condition Of Identifiability Hausman Specification Test The Method Of Indirect Least Squares (ILS): A Just Identified Equation The Method Of Two-Stage Least Squares (2SLS): An Over identified
Equation The Method Of Three-Stage Least Squares (3SLS) Using EViews The Granger Test Computer Application using EViews
Chapter 9 Dummy Variable Regression Models
The Nature of Dummy Variable Caution In the Use of Dummy Variables Analysis of Variance (ANOVA) Model Analysis of Covariance (ANCOVA) Models. Computer Application using EView
7|P a g e
CHAPTER 8 MULTIPLE REGRESSION ANALYSIS: THE PROBLEM OF INFERENCE

RESTRICTED LEAST SQUARES: TESTING LINEAR EQUALITY RESTRICTIONS
WALD Test: Wald test is used to test the validity of the linear restriction imposed on the parameters. There are occasions where economic theory may suggest that the coefficients in a regression model satisfy some linear equality restrictions.
For instance, consider the CobbDouglas production function
Y=KLe
Where Yi = output, Li = labor input, and Ki = capital input. Written in log form, the equation becomes
LnYi = 1 + 2 ln Ki + 3 ln Li + i (8.1)
Now if there are constant returns to scale (equi proportional change in output for an equiproportional change in the inputs), economic theory would suggest that
2 + 3 = 1
This is an example of a linear equality restriction. TESTING LINEAR RESTRICTIONS IN EVIEWS:
Step1: First Regress the Model (8.1) equation. Quick estimate equation write: log(Y) C log(K) log(L)
Dependent Variable: LOG(Y) Method: Least Squares Date: 08/21/13 Time: 13:41
Sample: 1960 2012 Included observations: 53 Coefficien t 10.94926 3.734688 -1.271933 0.996592 0.996456 0.025746 0.033143 120.2927 0.054050
Variable C LOG(K) LOG(L) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
Std. Error 0.420685 0.138653 0.214719
t-Statistic 26.02720 26.93548 -5.923718
Prob. 0.0000 0.0000 0.0000 18.36495 0.432456 -4.426138 -4.314612 7310.653 0.000000
Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)
Step2: First Regress the Model (8.1) equation Regression window View Coefficient Tests Wald-Coefficients Restrictions: Write: C(2) + C(3) = 1
Wald Test: Equation: Untitled
Test Statistic
Value
df
Probability
F-statistic Chi-square
319.2578 319.2578
(1, 50) 1
0.0000 0.0000
Null Hypothesis Summary:
Normalized Restriction (= 0)
Value
Std. Err.
-1 + C(2) + C(3)
1.462755
0.081865
Restrictions are linear in coefficients.
WALD TEST MANUALLY (USING EVIEWS):
Unrestricted model: lnYi = 0 + 2 ln Ki + 3 ln Li + i
STEP 1: First Regress the (8.1) equation. Quick estimate equation write: log(Y) C log(K) log(L) And Obtain RSSUR Dependent Variable: LOG(Y) Method: Least Squares Date: 08/21/13 Time: 13:41 Sample: 1960 2012 Included observations: 53 Coefficien t 10.94926 3.734688 -1.271933 0.996592 0.996456
Variable C LOG(K) LOG(L) R-squared Adjusted R-squared
Std. Error 0.420685 0.138653 0.214719
t-Statistic 26.02720 26.93548 -5.923718
Prob. 0.0000 0.0000 0.0000 18.36495 0.432456
Mean dependent var S.D. dependent var
10 | E C O N O M E T R I C S 2 T E R M P A P E R
S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
0.025746 0.033143 120.2927 0.054050
Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)
-4.426138 -4.314612 7310.653 0.000000
Restricted model: ln (Yi /Li) = 0 + 2 ln (Ki / Li) + i
STEP 2: First Regress the (8.2) equation. Quick estimate equation write: log(Y/L) C log(K/L) And Obtain RSSR
Dependent Variable: LOG(Y/L) Method: Least Squares Date: 08/21/13 Time: 13:48 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(K/L)
18.42422 5.936631
0.119150 0.170981
154.6300 34.72092
0.0000 0.0000
R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood
0.959412 0.958617 0.069277 0.244764 67.30666
Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic
14.30042 0.340546 -2.464402 -2.390052 1205.542
Durbin-Watson stat
0.035469
Prob(F-statistic)
0.000000
STEP 3: Apply the F Test of RSS Version. F = { (RSSR RSSUR) /M} / {RSSUR / (N-K)}
And compare it with F critical Values at m and (n-k) degrees of freedom; if | FCal | > | FCritical | then Restriction is invalid and vice versa
F = {0.244764 0.033143 / 1} / {0.033143 / (53-3)}

F = 319.69
The above example shows that F test is significant 1% level of significance.
TESTING FOR STRUCTURAL OR PARAMETER STABILITY OF REGRESSION MODELS: THE CHOW TEST When we use a regression model involving time series data, it may happen that there is a structural change in the relationship between the regressand Y and the regressors. By structural change, we mean that the values of the parameters of the model do not remain the same through the entire time period. USING E VIEWS:
Step 1: First Regress the Model (8.3) equation (with n = 34); Quick estimate equation write: Sav C Yd
Dependent Variable: SAV Method: Least Squares
Date: 08/21/13 Time: 14:08 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C YD
-15.73656 0.771388
1.331322 0.022667
-11.82025 34.03118
0.0000 0.0000
R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
0.957821 0.956994 0.880589 39.54728 -67.44468 0.024164
29.38251 4.246259 2.620554 2.694904 1158.121 0.000000
Step 2: First Regress the Model (8.3) equation. Regression window View Stability Tests Chow Breakpoint Test: Write Breakpoint Year in the Box: 1980
Chow Breakpoint Test: 1980
F-statistic Log likelihood ratio
303.2648 137.4620
Probability Probability
0.000000 0.000000
The above example shows that F test is significant 1% level of significance
MODEL DETERMINANTS OF LIFE EXPECTANCY IN PAKISTAN:

Y=B1+B2X1+B3X2++B4X3+B5X4+B6X5 X1: POPULATION X2: GDP X3: UNEMPLOYEMENT X4: URBAN POPULATION X5: HEALTH EXPENDITURE REGRESSION RESULTS:
Dependent Variable: Y Method: Least Squares Date: 08/19/13 Time: 13:57 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
R-squared Adjusted R-squared
0.997288 0.997000
Mean dependent var S.D. dependent var
58.49076 5.387357
S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
0.295091 4.092694 -7.334900 0.196325
Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)
0.503204 0.726256 3456.958 0.000000
CHAPTER 10 MULTICOLLINEARITY: WHAT HAPPENS REGRESSORS ARE CORRELATED? IF THE
THE NATURE OF MULTICOLLINEARITY The term multicollinearity is due to Ragnar Frisch. Originally it meant the existence of a perfect, or exact, linear relationship among some or all explanatory variables of a regression model. X1 + X2 . + Xk = 0 X1 + X2 . + Xk + i = 0 (10.1.1) (10.1.2)
Why does the classical linear regression model assume that there is no multicollinearity among the Xs? The reasoning is this: If multicollinearity is perfect in the sense of (10.1.1), the regression coefficients of the X variables are indeterminate and their standard errors are infinite. If multicollinearity is less than perfect, as in (10.1.2), the regression coefficients, although determinate, possess large standard errors (in relation to the coefficients themselves), which means the coefficients cannot be estimated with great precision or accuracy.
SOURCES OF MULTICOLLINEARITY There are several sources of multicollinearity. As Montgomery and Peck note, multicollinearity may be due to the following factors; 1. The data collection method employed, for example, sampling over a limited range of the values 2.
taken by the regressors in the population Constraints on the model or in the population being sampled. For example, in the regression of electricity consumption on income (X2) and house size (X3) there is a physical constraint in the population in that families with higher incomes generally have larger homes than families with lower incomes. Model specification, for example, is adding polynomial terms to a regression model, especially when the range of the X variable is small. An over-determined model. This happens when the model has more explanatory variables than the number of observations. This could happen in medical research where there may be a small number of patients about whom information is collected on a large number of variables.
3. 4.
An additional reason for multicollinearity, especially in time series data, may be that the regressors included in the model share a common trend, that is, they all increase or decrease over time.
PRACTICAL CONSEQUENCES OF MULTICOLLINEARITY In cases of near or high multicollinearity, one is likely to encounter the following consequences: a) Although BLUE, the OLS estimators have large variances and co variances, making precise estimation
difficult.
b) Because of consequence 1, the confidence intervals tend to be much wider, leading to the acceptance of the
zero null hypothesis (i.e., the true population coefficient is zero) more readily.
c) Also because of consequence 1, the t-ratio of one or more coefficients tends to be statistically insignificant. d) Although the t-ratio of one or more coefficients is statistically insignificant, R 2, the overall measure of
goodness of fit, can be very high.
e) The OLS estimators and their standard errors can be sensitive to small changes in the data.
DOING REGREESION USING EVIEWS FOR MULTICOLLINEARITY

DETECTION METHODS:
1. HIGH R2 BUT FEW SIGNIFICANT t- RATIO:
MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
STEP: First regress the above equation. Open the File containing data Quick estimate equation write: Y C X1 X2 X3 X4 X5 And check the R and t-ratios.
2
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
0.997288 0.997000 0.295091 4.092694 -7.334900
58.49076 5.387357 0.503204 0.726256 3456.958
Durbin-Watson stat
0.196325
Prob(F-statistic)
0.000000
The above regression results shows that R = 0.997288(high).X1, X2, X3, X4 are significant while X5 is insignificant. There may be multicollinearity here.
1.
HIGH PAIR WISE CORRELATION AMONG REGRESSORS:
STEP: First open the File containing data Quick Group Statistics Correlation Write: X1 X2 X3 X4 X5 X1 X1 X2 X3 X4 X5 1.000000 0.898710 0.900580 0.988885 -0.071599 X2 0.898710 1.000000 0.713399 0.863979 -0.222654 X3 0.900580 0.713399 1.000000 0.902906 -0.159905 X4 0.988885 0.863979 0.902906 1.000000 -0.053507 X5 -0.071599 -0.222654 -0.159905 -0.053507 1.000000
The above correlation
matrix shows that X1 and X4 are highly correlated.
2. EXAMINATION OF PARTIAL CORRELATION:

STEP: First regress the equations. Open the File containing data Quick estimate equation write: Y C X1, Y C X2, Y C X3, Y C X4, Y C X5 individually and compare their r2 with R2 of overall regression.
REGRESS Y ON X1:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1
46.13325 1.20E-07
0.681312 6.11E-09
67.71239 19.57344
0.0000 0.0000
R-squared
2
0.882521
Mean dependent var
58.49076
0.997288
REGRESS Y ON X2:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2
54.61037 7.42E-11
0.659574 8.68E-12
82.79642 8.554942
0.0000 0.0000
R-squared
2
0.589329
Mean dependent var
58.49076
0.997288
REGRESS Y ON X3:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X3
47.82694 3.221181
0.889198 0.245909
53.78662 13.09909
0.0000 0.0000
R-squared
2
0.770875
Mean dependent var
58.49076
0.997288
REGRESS Y ON X4:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X4
22.00695 1.241685
1.082998 0.036487
20.32039 34.03118
0.0000 0.0000
R-squared
2
0.957821
Mean dependent var
58.49076
0.997288
REGRESS Y ON X5:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X5
58.94157 -0.146728
5.968745 1.927387
9.875036 -0.076128
0.0000 0.9396
R-squared R
2
0.000114 0.997288
Mean dependent var
58.49076
The above regression shows data had no multicollinearity 3. AUXILLARY REGRESSIONS:

STEP: First regress the equations. Open the File containing data Quick estimate equation write: X2 C X3 X4, write: X3 C X2 X4, write: X4 C X2 X3 individually and compare the r2 with R2.
REGRESS X1 ON X2 X3 X4 X5:
Dependent Variable: X1 Method: Least Squares Date: 08/19/13 Time: 14:06 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2 X3 X4 X5
-1.22E+08 0.000186 4481648. 6366302. 4542256.
11339153 2.70E-05 1162156. 583130.7 1906829.
-10.78366 6.900419 3.856322 10.91745 2.382100
0.0000 0.0000 0.0003 0.0000 0.0212
R-squared R
2
0.989170 0.997288
Mean dependent var
1.03E+08
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X3 X4 X5
1.84E+11 2677.158 -2.16E+10 -8.45E+09 -2.88E+10
7.50E+10 387.9703 3.97E+09 3.94E+09 6.42E+09
2.451252 6.900419 -5.426820 -2.141991 -4.489499
0.0179 0.0000 0.0000 0.0373 0.0000
R-squared
2
0.910163
Mean dependent var
5.23E+10
0.997288
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X4 X5
1.445271 5.28E-08 -1.76E-11 -0.011486 -0.757352
2.266943 1.37E-08 3.25E-12 0.118091 0.189557
0.637542 3.856322 -5.426820 -0.097265 -3.995372
0.5268 0.0003 0.0000 0.9229 0.0002
R-squared
0.894134
Mean dependent var
3.310530
0.997288
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X5
18.57003 1.12E-07 -1.03E-11 -0.017156 -0.051514
0.745923 1.03E-08 4.82E-12 0.176382 0.267322
24.89536 10.91745 -2.141991 -0.097265 -0.192705
0.0000 0.0000 0.0373 0.9229 0.8480
R-squared
2
0.981090
Mean dependent var
29.38251
0.997288
Dependent Variable: X5
Method: Least Squares Date: 08/19/13 Time: 14:08 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4
2.736547 2.33E-08 -1.03E-11 -0.329525 -0.015007
1.448764 9.77E-09 2.29E-12 0.082477 0.077873
1.888884 2.382100 -4.489499 -3.995372 -0.192705
0.0650 0.0212 0.0000 0.0002 0.8480
R-squared
2
0.351577
Mean dependent var
3.072447
0.997288
The above regression shows data had no multicollinearity. REMEDIAL MEASURES: A. DROPPING A VARIABLE(S) AND SPECIFICATION BIAS :
STEP: First Regress the (10.1) equation without X1 (by assuming it causes Multicollinearity). Open the File containing data Quick estimate equation write: Y C X2 X4 And check the R2 and t-ratios
Dependent Variable: Y
Method: Least Squares Date: 08/19/13 Time: 14:12 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2 X3 X4 X5
9.097915 -3.84E-11 -0.751619 1.907816 -0.706191
1.536073 3.65E-12 0.157433 0.078995 0.258311
5.922841 -10.50413 -4.774216 24.15124 -2.733880
0.0000 0.0000 0.0000 0.0000 0.0087
0.987744 0.986722 0.620775 18.49733 -47.30811 0.340734
58.49076 5.387357 1.973891 2.159768 967.0998 0.000000
After dropping X1, regression shows that multicollinearity has been removed.
B. TRANSFORMATION OF VARIABLE: I m transforming model into lag form.
Dependent Variable: D(Y) Method: Least Squares Date: 08/19/13 Time: 14:18
Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C D(X1) D(X2) D(X3) D(X4) D(X5)
1.080539 -2.20E-07 -4.13E-12 0.002617 -0.482934 0.009456
0.091881 1.15E-08 1.19E-12 0.020385 0.278996 0.021666
11.76017 -19.15005 -3.461294 0.128386 -1.730969 0.436466
0.0000 0.0000 0.0012 0.8984 0.0902 0.6645
0.922727 0.914328 0.052655 0.127535 82.49098 1.059673
0.362255 0.179894 -2.941961 -2.716817 109.8591 0.000000
C. ADDITION OR NEW DATA.
CHAPTER 12 AUTOCORRELATION: WHAT HAPPENS IFTHE ERROR TERMS ARE CORRELATED?
THE NATURE OF AUTOCORRELATION The term autocorrelation may be defined as correlation between members of series of observations ordered in time [as in time series data] or space [as in cross-sectional data].In the regression context, the classical linear regression model assumes that such autocorrelation does not exist in the disturbances Ui. Tintner defines autocorrelation as lag correlation of a given series with itself, lagged by a number of time units, whereas he reserves the term serial correlation to lag correlation between two different series.
SOURCES OF AUTOCORRELATION
1. Inertia. A salient feature of most economic time series is inertia, or sluggishness. As is well known, time series such as GNP, price indexes, production, employment, and unemployment exhibit (business) cycles. 2. Specification Bias: Excluded Variables Case 3. Specification Bias: Incorrect Functional Form 4. Cobweb Phenomenon 5. Lags. 6. Manipulation of Data: Another source of manipulation is interpolation or extrapolation of data. 7. Data Transformation
PRACTICAL CONSEQUENCES OF AUTOCORRELATION In the presence of autocorrelation the usual OLS estimators, although linear, unbiased, and asymptotically (i.e., in large samples) normally distributed, are no longer minimum variance among all linear unbiased estimators. In short, they are not efficient relative to other linear and unbiased estimators. Put differently, they may not be BLUE. As a result, the usual, t, F, and may not be valid.
DOING REGREESION USING EVIEWS FOR AUTOCORRELATION
DETECTION METHOD: 1. GRAPHICAL METHOD:

Step 1: Regress the Model Open the File containing data Quick estimate equation Write: XCY Step 2: obtain the Residuals From the estimated equation result window Proc make residuals series (name) ok. Step 3: Plot these residuals and see the pattern for Autocorrelation. From step2 Quick Graph Scatter plot Write: r1(-1) r1
It is clear from the Graph that the residuals Rt and Rt-1 are serially correlated and positively correlated
2. RUNS TEST: MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Step 1: Regress the Model Open the File containing data Quick estimate equation Write: Y C X1 X2 X3 X4 X5
Step 2: obtain the Residuals From the estimated equation result window Proc make residuals series (name) ok.
Step 3: check the Runs by looking at changing sign of residuals, then find the interval by calculating mean variance the check whether Runs obtained lies in interval or not for clarity regarding Autocorrelation
(- - - - - - -, + + + + + + + + + + , - - - - - - - - - - - - - -,+, - - -, + + + + + + + + + +, - - - - -, + + ) R=8 N1=23 N2=29
MEAN(R) = 2N1N2 /N +1 = 1334/52+1 = 25.653 VARIANCE () = 2N1N2 (2N1N2 N) / N2 (N 1) = 1334*1282/2704*51
= 1710188/137904 = 12.4012
If the null hypothesis of randomness is sustainable, following the properties of the normal distribution, we should expect that PROB [E(R)-1.96 R E(R) +1.96] PROB [25.653-1.96(12.4012) 8 25.653+1.96(12.4012)] PROB [1.3466 8 49.959]
Obviously, this interval includes 8. So we can accept the null hypothesis that residuals do not contain autocorrelation.
3. DURBIN WATSON d TEST:
Step 1: Regress the Model Open the File containing data Quick estimate equation Write:
Y C X 2 X 3 X4 X5
In the regression results there is Durbin-Watson d statistics Dependent Variable: Y Method: Least Squares Date: 08/19/13 Time: 13:57 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
0.997288 0.997000 0.295091 4.092694 -7.334900 0.196325
58.49076 5.387357 0.503204 0.726256 3456.958 0.000000
4. A GENERAL TEST OF AUTOCORRELATION: THE BREUSCHGODFREY (BG) TEST MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Step 1: Regress the Model Open the File containing data Quick estimate equation equation Specification window Write:
Y C X 2 X 3 X4 X5
Step 2: from step 1, in estimated equation window View Residual Tests Serial Correlation LM Test Breusch-Godfrey Serial Correlation LM Test:
F-statistic Obs*R-squared
28.96111 42.88204
0.000000 0.000000
Test Equation: Dependent Variable: RESID Method: Least Squares Date: 08/19/13 Time: 14:44 Pre sample missing value lagged residuals set to zero.
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5 RESID(-1) RESID(-2) RESID(-3) RESID(-4) RESID(-5) RESID(-6)
-0.511900 -5.69E-09 1.71E-12 0.035050 0.027397 0.030173 0.772369 0.109866 0.067084 -0.015117 0.029055 -0.316267
0.658397 4.60E-09 1.19E-12 0.040369 0.034699 0.062743 0.154615 0.201820 0.201378 0.211764 0.218522 0.171752
-0.777494 -1.238405 1.433777 0.868241 0.789556 0.480893 4.995436 0.544377 0.333126 -0.071384 0.132961 -1.841416
0.4413 0.2226 0.1592 0.3903 0.4343 0.6331 0.0000 0.5891 0.7407 0.9434 0.8949 0.0728
R-squared
0.809095
Mean dependent var
-2.50E-15
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
0.757877 0.138045 0.781315 36.54857 1.489135
S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)
0.280545 -0.926361 -0.480257 15.79697 0.000000
REMEDIAL MEASURES: 1. NEWEYWEST CONSISTENT STANDARD ERRORS METHOD:
Step: Regress the Model Open the File containing data Quick estimate equation equation Specification window Write:
YCX2X3
In equation specification window options estimation option window tick Heteroscedasticity-consistent covariance NeweyWest click ok
Dependent Variable: Y Method: Least Squares Date: 08/21/13 Time: 11:22 Sample: 1960 2012 Included observations: 53 Newey-West HAC Standard Errors & Covariance (lag truncation=3)
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
2.390986 1.43E-08 2.39E-12 0.081656 0.133068 0.136790
-2.308518 -8.372859 -6.749603 -2.643598 20.05638 -1.193012
0.0254 0.0000 0.0000 0.0111 0.0000 0.2389
0.997288 0.997000 0.295091 4.092694 -7.334900 0.196325
58.49076 5.387357 0.503204 0.726256 3456.958 0.000000
Only Std. Error of coefficients are corrected for autocorrelation
2. GENERALIZED LEAST SQUARES (GLS) METHOD:

When is known: We use that rho to estimate the generalized difference equation through OLS then these results will be reliable and dont have autocorrelation problem. When is Unknown:
based on DurbinWatson d Statistic: 1 d/2

Step 1: Regress the Model Open the File containing data Quick estimate equation
Write:
Y C X1 X 2 X 3 X4 X5
And take Durbin Watson statistics and calculate rho (= p). Dependent Variable: Y Method: Least Squares Date: 08/19/13 Time: 13:57 Sample: 1960 2012 Included observations: 53
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
0.997288 0.997000 0.295091 4.092694 -7.334900 0.196325
58.49076 5.387357 0.503204 0.726256 3456.958 0.000000
1 0.196325/2 0.9018
Step 2: Regress the equation Open the File containing data Quick estimate equation Write: Y- 0.9018*d(Y) 1-0.9018 X1-0.9018*d(X1) X5-0.9018*(X5)
Dependent Variable: Y-0.9018*D(Y) Method: Least Squares Date: 08/21/13 Time: 11:30 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
1-0.9018 X1-0.9018*D(X1) X2-0.9018*D(X2) X3-0.9018*D(X3) X4-0.9018*D(X4) X5-0.9018*D(X5)
-54.11289 -1.17E-07 -1.71E-11 -0.235757 2.659116 -0.185270
13.94341 9.85E-09 2.88E-12 0.089920 0.071015 0.137451
-3.880893 -11.87465 -5.928562 -2.621862 37.44442 -1.347904
0.0003 0.0000 0.0000 0.0118 0.0000 0.1843
0.997261 0.996964 0.293827 3.971373 -6.909378
Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Durbin-Watson stat
58.39192 5.332351 0.496515 0.721658 0.187082
estimated from the Residuals:
t = t-1 + t
Y C X1 X 2 X 3 X4 X5
Obtain the Residuals, from the estimated equation result window Proc make Residuals series (name) ok
Step 2: Regress the residual on it lag to get rho to transform model in GDE Write: R1 R1 (-1) Dependent Variable: R1 Method: Least Squares Date: 08/21/13 Time: 11:32 Sample (adjusted): 1962 2012 Included observations: 51 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
D(R1)
0.356889
0.311125
1.147091
0.2568
0.024186 0.024186 0.268178 3.595959 -4.739484
0.010390 0.271481 0.225078 0.262957 0.133401
Step 3: Regress the equation
Open the File containing data Quick estimate equation Write: Y- 0.3568*d(Y) 1-0.3568 X1-0.3568*d(X1) X5-0.3568*(X5)
Dependent Variable: Y-.3568*D(Y) Method: Least Squares Date: 08/21/13 Time: 11:35 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
1-.3568 X1-.3568*D(X1) X2-.3568*D(X2) X3-.3568*D(X3) X4-.3568*D(X4) X5-.3568*D(X5)
-7.471330 -1.14E-07 -1.71E-11 -0.247995 2.635871 -0.198586
2.116689 9.48E-09 2.64E-12 0.090371 0.070181 0.146473
-3.529725 -12.06953 -6.461472 -2.744203 37.55831 -1.355786
0.0010 0.0000 0.0000 0.0086 0.0000 0.1818
0.997367 0.997081 0.282978 3.683530 -4.953129
58.58935 5.237607 0.421274 0.646418 0.146685
from Durbin two steps method: Step 1: Regress the equation given suggested by Durbin for his two step methods Open the File containing data Quick estimate equation
write: Y C X1 d(X1) X2 d(X2) X3 d(X3) X4 d(X4) X5 d(X5) d(Y) The coefficient of Yt-1 is estimated rho (=P) Dependent Variable: Y Method: Least Squares Date: 08/21/13 Time: 11:39 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 D(X1) X2 D(X2) X3 D(X3) X4 D(X4) X5 D(X5) D(Y)
-17.98627 -1.47E-07 -3.78E-07 -2.07E-11 4.08E-12 -0.213495 0.176647 3.177549 -0.389698 -0.037026 0.074601 3.333366
3.065724 1.22E-08 2.41E-07 2.75E-12 8.47E-12 0.091318 0.106567 0.142231 1.727353 0.130887 0.113507 1.153424
-5.866892 -12.00310 -1.570401 -7.517914 0.481797 -2.337915 1.657618 22.34072 -0.225604 -0.282886 0.657232 2.889976
0.0000 0.0000 0.1242 0.0000 0.6326 0.0245 0.1052 0.0000 0.8227 0.7787 0.5148 0.0062
0.998433 0.998002 0.231364 2.141171 9.152363 0.453925
58.71860 5.175647 0.109524 0.559812 2316.511 0.000000
= 3.3333
Step 2: Regress the equation (12.4) Open the File containing data Quick estimate equation Write: Y- 3.3333*d(Y) 1-3.333 X1-3.3333*d(X1) X5-3.3333*(X5)
Dependent Variable: Y-3.333*D(Y) Method: Least Squares Date: 08/21/13 Time: 11:43 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
1-3.333 X1-3.333*D(X1) X2-3.333*D(X2) X3-3.333*D(X3) X4-3.333*D(X4) X5-3.333*D(X5)
3.158922 -1.40E-07 -9.24E-12 -0.060601 2.759547 -0.032140
0.687286 1.09E-08 3.84E-12 0.055325 0.086447 0.061097
4.596224 -12.87266 -2.405610 -1.095376 31.92176 -0.526048
0.0000 0.0000 0.0202 0.2791 0.0000 0.6014
0.994743 0.994172 0.439444 8.883108 -27.84041
57.51121 5.756341 1.301554 1.526698 0.369519
BerenbluttWebb test g-statistics for estimated =1: To test the hypothesis that 1. The test statistic they use is called the g-statistic, which is defined as follows: g = et2 / t2
Step 1: Regress the equation Open the File containing data Quick estimate equation Write: Y C X1 X2 X3 X4 X5 Obtain residuals, t
2
Step 2: Regress the equation First difference Open the File containing data Quick estimate equation Write: D(Y) C D(X10 D(X2) D (X3) D(X4) D(X5) Obtain residuals, et
2
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3
-5.519635 -1.20E-07 -1.61E-11 -0.215865
1.350874 9.29E-09 2.45E-12 0.085649
-4.085972 -12.86161 -6.580235 -2.520339
0.0002 0.0000 0.0000 0.0152
X4 X5
2.668870 -0.163192
0.070082 0.129846
38.08228 -1.256813
0.0000 0.2150
R-squared Adjusted R-squared S.E. of regression Sum squared resid
0.997288 0.997000 0.295091 4.092694
Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion
58.49076 5.387357 0.503204 0.726256
Dependent Variable: D(Y) Method: Least Squares Date: 08/21/13 Time: 16:29 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C D(X1) D(X2) D(X3) D(X4) D(X5)
1.080539 -2.20E-07 -4.13E-12 0.002617 -0.482934 0.009456
0.091881 1.15E-08 1.19E-12 0.020385 0.278996 0.021666
11.76017 -19.15005 -3.461294 0.128386 -1.730969 0.436466
0.0000 0.0000 0.0012 0.8984 0.0902 0.6645
0.922727 0.914328 0.052655 0.127535
0.362255 0.179894 -2.941961 -2.716817
g = 0.0311
We find that dL=1.364 and dU =1.590 (5 percent level)
It is clear that g statistics lies in the 0 dL range we can take first difference by assuming =1.
CHAPTER 11 HETEROSCEDASTICITY: WHAT HAPPENS IF THE ERROR VARIANCE IS NONCONSTANT?
THE NATURE OF HETEROSCEDASTICITY A critical assumption of the classical linear regression model is that the disturbances i have all the same variance, 2 if this assumption is not satisfied, there is heteroscedasticity. Hence, there is heteroscedasticity. E i = i2 Notice the subscript of i2, which reminds us that the conditional variances of i (= conditional variances of Yi) are no longer constant
SOURCES OF HETEROSCEDASTICITY
1. Following the error-learning models, as people learn, their errors of behavior become smaller over time. In this case, i2 is expected to decrease. 2. As data collecting techniques improve, i2 is likely to decrease. 3. Heteroscedasticity can also arise as a result of the presence of outliers. An outlying observation, or outlier, is an observation that is much different (either very small or very large) in relation to the observations in the sample. 4. Another source of heteroscedasticity arises from violating Assumption 9 of CLRM, namely, that the regression model is correctly specified. 5. Another source of heteroscedasticity is skewness in the distribution of one or more regressors included in the model. 6. Other sources of heteroscedasticity: As David Hendry notes, heteroscedasticity can also arise because of (1) incorrect data transformation (e.g., ratio or first difference transformations) and (2) incorrect functional form (e.g., linear versus log linear models).
PRACTICAL CONSEQUENCES OF HETEROSCEDASTICITY In the presence of heteroscedasticity the usual OLS estimators, although linear, unbiased, and asymptotically (i.e., in large samples) normally distributed, are no longer minimum variance among all linear unbiased estimators. In short, they are not efficient relative to other linear and unbiased estimators. Put differently, they may not be BLUE. As a result, the usual, t, F, and may not be valid.
DOING REGREESION USING EVIEWS FOR HETEROSCEDASTICITY
DETECTION OF HETEROSCEDASTICITY: 1. GRAPHICAL METHOD: MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Step 1: Regress the Model Open the File containing data Quick estimate equation Write: Y C X1 X2 X3 X4 X5
Step 2: obtain the Residuals From the estimated equation result window Proc make residuals series (name) ok.
Step 3: Plot these residuals squares and X1, X2, X3, X4, X5 & individually and see the pattern for Hetroscedasticity. Y From step2 Quick Graph Scatter plot Write: X1 R1^2, X2 R1^2, X3 R1^2, X4 R1^2, X5 R1^2
It is clear from the graph that there is such pattern seen that causes hetroscedasticity.
2. FORMAL METHODS:
a. PARK TEST:
Y C X1X 2
And obtain the Residuals from the estimated equation result window Proc make Residuals series (name) ok.
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
0.997288 0.997000 0.295091 4.092694
58.49076 5.387357 0.503204 0.726256
Log likelihood Durbin-Watson stat
-7.334900 0.196325
F-statistic Prob(F-statistic)
3456.958 0.000000
Lni = 1++ 2lnXi+i
Step 2: Regress the Equation (11.1) suggested by Park. Open the File containing data Quick estimate equation Write: log(r1^2) c log(x1) log(r1^2) c log(x2) log(r1^2) c log(x3) log(r1^2) c log(x4) log(r1^2) c log(x5) And check the significance of coefficient of explanatory variable
Dependent Variable: LOG(R1^2) Method: Least Squares Date: 08/21/13 Time: 12:07 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X1)
39.65786 -2.361010
10.25252 0.557693
3.868109 -4.233531
0.0003 0.0001
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X2)
17.17790 -0.866555
5.018112 0.207702
3.423181 -4.172101
0.0012 0.0001
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X3)
-1.586677 -1.935249
0.639137 0.532505
-2.482532 -3.634238
0.0164 0.0007
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X4)
19.37869 -6.848214
5.598606 1.657294
3.461342 -4.132166
0.0011 0.0001
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X5)
-5.142800 1.264550
2.376507 2.120873
-2.164016 0.596240
0.0353 0.5537
If 2s turns out to be statistically significant at 1%, it would suggest that heteroscedasticity is present in the data in the each case.
b. GLESJER TEST: MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Step 1: Regress the Model Open the File containing data Quick estimate equation
Write:
Y C X1X 2
And obtain the Residuals from the estimated equation result window Proc make Residuals series (name) ok
= 1+2+Xi+i
Step 2: Regress the Equation (11.1a) suggested by Glesjer. Open the File containing data Quick estimate equation Write: abs(r1) c x1 abs(r1) c x2 abs(r1) c x3 abs(r1) c x4 abs(r1) c x5 And check the significance of coefficient of explanatory variable
Dependent Variable: ABS(R1) Method: Least Squares Date: 08/21/13 Time: 12:16 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1
0.462368 -2.31E-09
0.051704 4.60E-10
8.942531 -5.020498
0.0000 0.0000
Dependent Variable: ABS(R1)
Method: Least Squares Date: 08/21/13 Time: 12:16 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2
0.290397 -1.30E-12
0.029296 3.82E-13
9.912631 -3.408431
0.0000 0.0013
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X3
0.423221 -0.060295
0.050627 0.013883
8.359624 -4.343235
0.0000 0.0001
Variable
Coefficient
Std. Error
t-Statistic
Prob.
0.914931
0.137492
6.654407
0.0000
X4
-0.023500
0.004612
-5.094806
0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X5
0.123370 0.031866
0.185112 0.059835
0.666459 0.532554
0.5082 0.5967
If 2s turns out to be statistically significant at 1%, it would suggest that heteroscedasticity is present in the data in each case.
= 1+2+Xi+ i
Step 2: Regress the Equation (11.1b) suggested by Glesjer. Open the File containing data Quick estimate equation Write: abs(r1) c x1^0.5 abs(r1) c x2^0.5 abs(r1) c x3^0.5 abs(r1) c x4^0.5
abs(r1) c x5^0.5 And check the significance of coefficient of explanatory variable
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1^.5
0.697759 -4.76E-05
0.093902 9.19E-06
7.430750 -5.181553
0.0000 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2^.5
0.383647 -8.01E-07
0.041604 1.80E-07
9.221475 -4.443441
0.0000 0.0000
Dependent Variable: ABS(R1) Method: Least Squares Date: 08/21/13 Time: 12:18 Sample (adjusted): 1961 2012
Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X3^.5
0.633219 -0.230573
0.090155 0.049249
7.023649 -4.681793
0.0000 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X4^.5
1.599442 -0.254295
0.269902 0.049674
5.926016 -5.119270
0.0000 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X5^.5
0.004676 0.123829
0.366416 0.209160
0.012761 0.592029
0.9899 0.5565
= 1+2+1/Xi+ i
Step 2: Regress the Equation (11.1c) suggested by Glesjer. Open the File containing data Quick estimate equation Write: abs(r1) c 1/x1 abs(r1) c 1/x2 abs(r1) c 1/x3 abs(r1) c 1/x4 abs(r1) c 1/x5 And check the significance of coefficient of explanatory variable
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X1
-0.010629 20352326
0.048537 3925383.
-0.218995 5.184800
0.8275 0.0000
Dependent Variable: ABS(R1) Method: Least Squares
Date: 08/21/13 Time: 12:20 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X2
0.135861 1.42E+09
0.026436 2.98E+08
5.139347 4.761296
0.0000 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X3
0.009848 0.577436
0.044574 0.110386
0.220930 5.231046
0.8260 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
-0.451516
0.132770
-3.400733
0.0013
1/X4
19.46190
3.801832
5.119086
0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X5
0.357066 -0.410269
0.180871 0.541472
1.974150 -0.757693
0.0539 0.4522
= 1+2+1/Xi+ i
Step 2: Regress the Equation (11.1d) suggested by Glesjer. Open the File containing data Quick estimate equation Write: abs(r1) c 1/x1^0.5 abs(r1) c 1/x2^0.5 abs(r1) c 1/x3^0.5 abs(r1) c 1/x4^0.5 abs(r1) c 1/x5^0.5 And check the significance of coefficient of explanatory variable
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X1^0.5
-0.245656 4471.569
0.090575 848.7235
-2.712173 5.268582
0.0091 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X2^0.5
0.047878 25715.18
0.037971 4894.805
1.260883 5.253565
0.2132 0.0000
Dependent Variable: ABS(R1) Method: Least Squares Date: 08/21/13 Time: 12:22
Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X3^0.5
-0.199286 0.713478
0.083843 0.138597
-2.376886 5.147859
0.0213 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X4^0.5
-1.135887 7.318108
0.265139 1.426138
-4.284128 5.131418
0.0001 0.0000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X5^0.5
0.475984 -0.443667
0.362154 0.629237
1.314313 -0.705088
0.1947 0.4840
If 2s turns out to be statistically significant at 1%, it would suggest that heteroscedasticity is present in the data in each case
c. WHITES GENERAL HETEROSCEDASTICITY TEST: MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Y C X1X 2 X3 X4 X5
Step 2: applying Whites General Hetroscedasticity Test In estimated equation in step1 view Residuals Tests White Heteroskedasticity Test (cross terms) or White Heteroskedasticity Test (no cross term)
White Heteroskedasticity Test: no cross term
5.982381 31.13871
0.000014 0.000557
As n*R2 (=31.13871) > the critical chi-square (=3.94) value at the 5% level of significance, the conclusion is that there is heteroscedasticity
White Heteroskedasticity Test: cross term
3.782162 37.24425
0.000405 0.010937
As n*R2 (=37.24425) > the critical chi-square (=10.85) value at the 5% level of significance, the conclusion is that there is heteroscedasticity
d. KOENKERBASSETT (KB) TEST:
Y C X1X 2 X3 X4 X5
I2= 1+2I2 +i
Step 2: Regress the (11.6a) equation Open the File containing data Quick estimate equation Write: in R1^2 C (Y_CAP)^2 Where I are the estimated values from the model (11.6). The null hypothesis is that 2= 0. If this is not rejected, then one could conclude that there is no heteroscedasticity. The null hypothesis can be tested by the usual t test or the F test,
Dependent Variable: R1^2 Method: Least Squares Date: 08/21/13 Time: 12:44 Sample (adjusted): 1961 2012 Included observations: 52 after adjustments
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C (Y_CAP)^2
0.414590 -9.74E-05
0.072768 2.07E-05
5.697453 -4.714133
0.0000 0.0000
0.307700 0.293854 0.087643 0.384066 53.82797 0.407608
0.076373 0.104297 -1.993383 -1.918336 22.22305 0.000020
REMEDIAL MEASURES:
WHEN i2 IS KNOWN: THE METHOD OF WEIGHTED LEAST SQUARES: If i2 is known, the most straightforward method of correcting heteroscedasticity is by
Means of weighted least squares, for the estimators thus obtained are BLUE Yi = 1 + 2 X1i + 3 X2i + 4 X3i + 5X4i + 6X5i +i Yi/ = 1 (1/)+ 2 X1i /+ 3 X2i/+ 4 X3i/ +
Where i are the standard deviations
/+6X5i/+ i WEIGHTED LEAST SQUARE
WHEN i2 IS NOT KNOWN: Whites Heteroscedasticity-Consistent Variances and Standard Errors. MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i
Y C X1X 2 X3 X4 X5
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846
-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813
0.0002 0.0000 0.0000 0.0152 0.0000 0.2150
Y C X1X 2 X3 X4 X5
In equation specification window options estimation option window tick Heteroscedasticity-consistent covariance Whites click ok
Dependent Variable: Y Method: Least Squares Date: 08/22/13 Time: 14:14 Sample: 1960 2012 Included observations: 53 White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X3 X4 X5
-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192
1.416856 8.29E-09 1.64E-12 0.056197 0.078266 0.106943
-3.895694 -14.42540 -9.811624 -3.841198 34.09986 -1.525967
0.0003 0.0000 0.0000 0.0004 0.0000 0.1337
As we can see that in case of heteroscedasticity the OLS standard errors of slope coefficient of X1 X2 X3 are over estimated and X4 X5 are under estimated. And intercept was under estimated. After applying Whites Hetroscedasticity-Consistent Variances and Standard Errors remedial procedure they were now corrected for hetroscedasticity.
Plausible Assumptions about Heteroscedasticity Pattern :
Assumption 1: The error variance is proportional to XI2: E(i2)=2Xi2
We may transform by dividing the original model through by Xi
Transforming using X1: Yi = 1 (1/X1) + 2 + 3 X2i/X1 + 4 X3i/X1 + +6X5i/X1 +i TRANSFORMED MODEL
Step 1: Regress the equation (11.7a) Open the File containing data Quick estimate equation equation specification window Write: Y/X1 C 1/X1 X2/X1 X3/X1 X4/X1 X5/X1
Dependent Variable: Y/X1 Method: Least Squares Date: 08/22/13 Time: 14:31 Sample: 1960 2012 Included observations: 53 White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X1 X2/X1 X3/X1 X4/X1 X5/X1
-1.35E-07 -8.939349 -1.65E-11 -0.360089 2.859217 -0.206358
1.30E-08 1.641135 3.01E-12 0.097640 0.091651 0.123891
-10.36097 -5.447055 -5.480641 -3.687939 31.19696 -1.665644
0.0000 0.0000 0.0000 0.0006 0.0000 0.1024
R-squared
0.999641
Mean dependent var
6.51E-07
Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat
0.999603 4.34E-09 8.85E-16 948.5141 0.188567
S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)
2.18E-07 -35.56657 -35.34352 26205.57 0.000000
Similarly for X2 X3 X4 and X5:
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1/X2 1/X2 X3/X2 X4/X2 X5/X2
-1.60E-11 -1.80E-07 -15.75433 -0.400503 3.241575 -0.141550
1.20E-11 2.77E-08 2.283040 0.187234 0.140842 0.103679
-1.337315 -6.486804 -6.900594 -2.139050 23.01573 -1.365271
0.1876 0.0000 0.0000 0.0377 0.0000 0.1787
0.999966 0.999962 2.12E-11 2.11E-20 1230.651
3.39E-09 3.44E-09 -46.21324 -45.99019 274036.8
Durbin-Watson stat
0.158014
Prob(F-statistic)
0.000000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1/X3 X2/X3 1/X3 X4/X3 X5/X3
-0.308845 -1.49E-07 -1.40E-11 -10.39333 2.948408 -0.215436
0.117205 1.46E-08 3.08E-12 1.716490 0.095104 0.120129
-2.635089 -10.18503 -4.564873 -6.054987 31.00182 -1.793374
0.0114 0.0000 0.0000 0.0000 0.0000 0.0793
0.999735 0.999707 0.137977 0.894776 32.95532 0.192584
20.99458 8.062351 -1.017182 -0.794130 35499.75 0.000000
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1/X4 X2/X4 X3/X4 1/X4 X5/X4
2.734611 -1.25E-07 -1.64E-11 -0.255739 -6.702676 -0.189060
0.080073 9.49E-09 1.99E-12 0.068636 1.430325 0.115224
34.15149 -13.13093 -8.252539 -3.726028 -4.686120 -1.640812
0.0000 0.0000 0.0000 0.0005 0.0000 0.1075
0.990734 0.989748 0.011333 0.006037 165.4213 0.185140
2.006037 0.111929 -6.015897 -5.792845 1005.009 0.000000
Dependent Variable: Y/X5 Method: Least Squares Date: 08/22/13 Time: 14:35 Sample: 1960 2012
Included observations: 53 White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1/X5 X2/X5 X3/X5 X4/X5 1/X5
-0.107515 -1.19E-07 -1.56E-11 -0.213828 2.657934 -5.458239
0.114543 7.82E-09 1.60E-12 0.057516 0.074921 1.338748
-0.938644 -15.21492 -9.744259 -3.717711 35.47637 -4.077122
0.3527 0.0000 0.0000 0.0005 0.0000 0.0002
0.999139 0.999047 0.094632 0.420893 52.94151 0.202398
19.34730 3.065848 -1.771378 -1.548326 10906.54 0.000000
Notice that in the transformed regression the intercept term 2 is the slope coefficient in the original equation and the slope coefficient 1 is the intercept term in the original model. Therefore, to get back to the original model we shall have to multiply the estimated transformed model by Xi.
Assumption 2: The error variance is proportional to Xi. The square root transformation: E (i2) =2Xi
Transforming using X1:
Yi /X1= 1 (1/X1) + 2X1 + 3 X2i/X1 + 4 X3i/X1 +

TRANSFORMED
+6X5i/X1 +i
Step 1: Regress the equation (11.8a) Open the File containing data Quick estimate equation equation specification window Write:
Y/x1^0.5 c 1/x1^0.5 x2/x1^0.5 x3/x1^0.5 x4/x1^0.5 x5/x1^0.5
Dependent Variable: Y/X1^0.5 Method: Least Squares Date: 08/22/13 Time: 14:42 Sample: 1960 2012 Included observations: 53 White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C 1/X1^0.5 X2/X1^0.5 X3/X1^0.5 X4/X1^0.5 X5/X1^0.5
-0.003423 -3.483622 -2.95E-11 -0.356717 3.378221 -0.166805
0.000293 1.212010 1.99E-12 0.062820 0.131543 0.123406
-11.69658 -2.874252 -14.88087 -5.678385 25.68143 -1.351681
0.0000 0.0061 0.0000 0.0000 0.0000 0.1829
0.998082 0.997878 3.45E-05 5.58E-08 472.5751 0.287508
0.006034 0.000748 -17.60661 -17.38355 4892.212 0.000000
Similarly do for X2 X3 X4 and X5. Note an important feature of the transformed model: It has no intercept term. Therefore, one will have to use the regression-through-the-origin model to Estimate 1 and 2. Having run (11.8), one can get back to the original model Simply by multiplying transformed model by Xi. Assumption 3: The error variance is proportional to the square of the mean value of Y. E (i2)=2E(Yi)2 Equation postulates that the variance of i2 is proportional to the square of the expected Value of Y. Therefore, if we transform the original equation as follows: Yi /E (Yi) = 1 (1/ E (Yi)) + 2X1/E (Yi) +3 X2i/ E (Yi) + 4 X3i/ E (Yi) + +i +6X5i/ E (Yi)
The transformation (11.9) is, however, in operational because E (Yi) depends on 1 and 2, which are unknown. Of course, we know, which is an estimator of E (Yi) Transforming using: i Yi /i = 1 (1/ i) + 2X1i/ i + 3 X2i/ i + 4 X3i/ i + +6X5i/ i + i
Step: Regress the equation (11.8a) Open the File containing data Quick estimate equation equation specifications window Write: Y/Y_CAP 1/Y_CAP X1/Y_CAP X2/Y_CAP X3/Y_CAP X4/Y_CAP X5/Y_CAP
Dependent Variable: Y/Y_CAP Method: Least Squares Date: 08/22/13 Time: 14:55 Sample: 1960 2012 Included observations: 53
White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
1/Y_CAP X1/Y_CAP X2/Y_CAP X3/Y_CAP X4/Y_CAP X5/Y_CAP
-6.465594 -1.24E-07 -1.62E-11 -0.237086 2.721034 -0.182424
1.412492 9.12E-09 1.85E-12 0.063967 0.078917 0.112227
-4.577437 -13.58257 -8.803458 -3.706393 34.47970 -1.625497
0.0000 0.0000 0.0000 0.0006 0.0000 0.1107
0.013722 -0.091201 0.005434 0.001388 204.3804
0.999988 0.005202 -7.486055 -7.263003 0.186775
Assumption 4: A log transformation such as: lnYi = 1 + 2 lnXi+ i very often reduces heteroscedasticity. Log Transforming: lnYi = 1 + 2 lnX1i + 3 lnX2i + 4 lnX3i + 5lnX4i+6lnX5i+i
Step: Regress the equation (11.8c) Open the File containing data Quick estimate equation equation specifications window Write: log(y) c log(x1) log(x2) log(x3) log(x4) log(x5)
Dependent Variable: LOG(Y) Method: Least Squares Date: 08/22/13 Time: 15:00 Sample: 1960 2012 Included observations: 53 White Heteroskedasticity-Consistent Standard Errors & Covariance
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C LOG(X1) LOG(X2) LOG(X3) LOG(X4) LOG(X5)
5.341832 -0.397072 -0.040926 0.020821 2.058282 0.037678
0.443418 0.050292 0.010557 0.008782 0.171616 0.012541
12.04694 -7.895290 -3.876748 2.370920 11.99351 3.004484
0.0000 0.0000 0.0003 0.0219 0.0000 0.0043
0.988388 0.987152 0.010827 0.005509 167.8449 0.393330
4.064502 0.095517 -6.107355 -5.884303 800.0900 0.000000
This result arises because log transformation compresses the scales in which the variables are measured, thereby reducing a tenfold difference between two values to a twofold difference. To conclude our discussion of the remedial measures, we reemphasize that all the transformations discussed previously are ad hoc; we are essentially speculating
about the nature of i2. Which of the transformations discussed previously will work will depend on the nature of the problem and the severity of heteroscedasticity.
CHAPTERS 18 TO 20 SIMULTANEOUS REGRESSION MODELS
THE NATURE OF SIMULTANEOUS-EQUATION MODELS In contrast to single-equation models, in simultaneous-equation models more than one dependent, or endogenous, variable is involved, necessitating as many equations as the number of endogenous variables. A unique feature of simultaneous-equation models is that the endogenous variable (i.e., regressand) in one equation may appear as an explanatory variable (i.e., regressor) in another equation of the system. Y1i = 10 + 12Y2i + 11 X1i + u1i Y2i = 20 + 21Y1i + 21 X1i + u2i Where Y1 and Y2 are mutually dependent, or endogenous, variables and X1 is an exogenous variable and where u1 andu2 are the stochastic disturbance terms, the variables Y1 and Y2 are both stochastic. Therefore, unless it can be shown that the stochastic explanatory variable Y2 in (18.1) is distributed independently of u1 and the stochastic explanatory variable Y1in (18.2) is distributed independently of u2, application of the classical OLS to these equations individually will lead to inconsistent estimates.
As a consequence, such an endogenous explanatory variable becomes stochastic and is usually correlated with the disturbance term of the equation in which it appears as an explanatory variable. In this situation the classical OLS method may not be applied because the estimators thus obtained are not consistent, that is, they do not converge to their true population values no matter how large the sample size. Since simultaneous-equation models are used frequently, especially in econometric models, alternative estimating techniques have been developed by various authors.
DOING REGREESION USING EVIEWS SIMULTENENOUS MODELS
MODEL: Yi = 1 + 2 X1i + 3 X2i +4X4i + 5X5i +1i X1i= 1+2Yi+3X2i+4X3i+ 2i
Where Yi and X1i are mutually dependent, or endogenous, variables X2i, X3i, X4i and X5i are an exogenous variable and where 1i and 2i are the stochastic disturbance terms, the variables Yi and X1i are both stochastic. Therefore, unless it can be shown that the stochastic explanatory variable Yi is distributed independently of 1i and the stochastic explanatory variable X1i is distributed independently of 2i, application of the classical OLS to these equations individually will lead to inconsistent estimates.
THE IDENTIFICATION PROBLEM RULES FOR IDENTIFICATION THE ORDER CONDITION OF IDENTIFIABILITY A necessary (but not sufficient) condition of identification, known as the order condition, May be stated in two different but equivalent ways as follows (the necessary as well as
Sufficient condition of identification will be presented shortly):
MKG-1
M = numbers of variables in the simultaneous model
i
K = numbers of variables in the specific equation
G = numbers of endogenous variables or equations
FOR EQUATION 1: 6521 11 FOR EQUATION 2: 6 4 2-1 21
Hence equation 1 is just identified and equation 2 is over identified. As we know equation 1 is just identified, is estimated using ILS and 2SLS and equation 2 is over identified, is estimated using 2SLS method.
THE RANK CONDITION FOR IDENTIFIABILITY:
Here we can not apply the rank condition there are only two simultaneous equations only
HAUSMAN SPECIFICATION TEST It is also called test of endogeneity or simultaneity problem

EVIEWS: Step 1: Regress the equation 1. Run Yi
on X1 X2 X3 X4 X5
Open the File containing data Quick estimate equation
Write: Y C X1 X2 X4 X5 And obtain the Residuals from the estimated equation result window Proc make Residuals series (name) ok
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X2 X3 X4 X5
9.097915 -3.84E-11 -0.751619 1.907816 -0.706191
1.536073 3.65E-12 0.157433 0.078995 0.258311
5.922841 -10.50413 -4.774216 24.15124 -2.733880
0.0000 0.0000 0.0000 0.0000 0.0087
0.987744 0.986722 0.620775 18.49733 -47.30811 0.340734
58.49076 5.387357 1.973891 2.159768 967.0998 0.000000
Step 2: Regress the equation 2 using residual of step 1 as explanatory variable
Open the File containing data Quick estimate equation Write: X1 C Y X2 X3 R1 Now check the significance of this residual coefficient if it exceeds critical then we can say both the equation are simultaneous
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C Y X2 X3 RESID01
-1.53E+08 3794853. 0.000291 5855713. -10309118
10008321 203298.8 1.31E-05 671827.1 797193.2
-15.31155 18.66638 22.24633 8.716101 -12.93177
0.0000 0.0000 0.0000 0.0000 0.0000
0.994332 0.993859 3315248. 5.28E+14 -868.3221 0.634807
1.03E+08 42305993 32.95555 33.14143 2104.972 0.000000
As residual is significant 1% level of significance; w can conclude that both equation are simultaneous
Similarly for equation 2 Repeat step 1 now.

Step 2: Regress the equation 1 using residual of step 1 as explanatory variable Open the File containing data Quick estimate equation Write: y c x1 x2 x4 x5 r1 Now check the significance of this residual coefficient if it exceeds critical then we can say both the equation are simultaneous
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X1 X2 X4 X5 RESID01
-0.811293 -9.78E-08 -1.70E-11 2.397581 -0.049479 0.362655
1.899331 1.21E-08 2.24E-12 0.101523 0.107277 0.101890
-0.427146 -8.112440 -7.562896 23.61622 -0.461224 3.559284
0.6712 0.0000 0.0000 0.0000 0.6468 0.0009
0.997575 0.997317 0.279035 3.659450 -4.369795
58.49076 5.387357 0.391313 0.614365 3867.342
Durbin-Watson stat
0.248237
Prob(F-statistic)
0.000000
As residual is significant 1% level of significance; w can conclude that both equation are simultaneous
CHAPTER 9 DUMMY VARIABLE REGRESSION MODELS
THE NATURE OF DUMMY VARIABLE Now, we consider models that may involve not only ratio scale variables but also nominal scale variables. Such variables are also known as indicator variables, categorical variables, qualitative variables, or dummy variables. Since dummy usually indicate the presence or absence of a quality or an attribute, such as male or female, black or white, Catholic or non-Catholic, Democrat or Republican, they are essentially nominal scale variables. Variables that assume
such 0 and 1 values are called dummy variables. Dummy variables can be incorporated in regression models just as easily as quantitative variables. As a regression model may contain regressors that are all exclusively dummy, or qualitative, in nature. Such models are called Analysis of Variance (ANOVA) models.
CAUTION IN THE USE OF DUMMY VARIABLES
1. For each qualitative regressor the number of dummy variables introduced must be one less than the categories of that variable. If you do not follow this rule, you will fall into what is called the dummy variable trap, that is, the situation of perfect co linearity or perfect multi-co linearity, if there is more than one exact relationship among the variables. 2. The category for which no dummy variable is assigned is known as the base, benchmark, control, comparison, reference, or omitted category. And all comparisons are made in relation to the benchmark category. 3. The intercept value (1) represents the mean value of the benchmark category. In Equation (9.1), the benchmark category is the Rural-North region. 4. The coefficients attached to the dummy variables in (9.1) are known as the differential intercept coefficients because they tell by how much the value of the intercept that receives the value of 1 differs from the intercept coefficient of the benchmark category. Which is a better method of introducing a dummy variable: (1) intro-duce a dummy for each category and omit the intercept term or (2) include the intercept term and introduce only (m1) dummies, where m is the number of categories of the dummy variable? Regression models containing an admixture of quantitative and qualitative variables are called analysis of covariance (ANCOVA) models. ANCOVA models are an extension of the ANOVA models in that they provide a method of statistically controlling the effects of quantitative regressors, called covariates or control variables, in a model that includes both quantitative and qualitative, or dummy, regressors.
DOING REGREESION ON DUMMY MODELS USING EVIEWS
ANALYSIS OF VARIANCE (ANOVA) MODEL: MODEL: Yi = 1 + 2D2i + 1i Where Yi is for life expectancy D2 = 0 for male D2 = 1 for female
First Regress the (9.1) equation Open the File containing data Quick estimate equation Write: y c d2
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C D2
58.44163 0.100143
1.046867 1.494661
55.82529 0.067001
0.0000 0.9468
0.000088 -0.019518 5.439678 1509.095 -163.9515 0.005902
58.49076 5.387356 6.262322 6.336673 0.004489 0.946843
When D2=0 Yi = 1 + 2D2i + 1i Yi = 58.44163 + 0.100143(0) + 1i Yi = 58.44163 mean life expectancy for male
When D2 = 1 Yi = 1 + 2D2i + 1i Yi = 58.44163 + 0.100143(1) + 1i Yi = 58.541773 mean life expectancy for female
ANALYSIS OF COVARIANCE (ANCOVA) MODEL MODEL: Yi = 1 + 2X3i + 3D2i + 1i Where Yi is life expectancy X3i is unemployment
D2 = 1 for Urban D2 = 0 for Other (Rural)
First regress the equation Open the File containing data Quick estimate equation Write: Y C X3 D2
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C X3 D2
47.80963 3.221096 0.035865
0.963399 0.248355 0.722599
49.62598 12.96970 0.049633
0.0000 0.0000 0.9606
0.770886 0.761722 2.629770 345.7845 -124.9051 0.254920
58.49076 5.387356 4.826608 4.938134 84.11621 0.000000
When D2 = 0 Yi = 47.80963+ 3.221096X3i+ 1i mean life expectancy function for rural When D2 = 1 Yi = 47.80963 + 3.221096X3i + 0.035865(1) + 1i Yi = 47.845495 + 3.221096X3i + 1i mean life expectancy function for urban

Applying Regression Using e Views (With Commands)

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Applying Regression Using e Views (With Commands)

Enviado por

Direitos autorais:

Formatos disponíveis

1|E C O N O M E T R I C S 2 T E R M P A P E R

APPLIED ECONOMICS RESEARCH CENTRE

Computer Application using EViews

Chapters 18 To 20 Simultaneous Regression Models

Chapter 9 Dummy Variable Regression Models

CHAPTER 8 MULTIPLE REGRESSION ANALYSIS: THE PROBLEM OF INFERENCE

For instance, consider the CobbDouglas production function

Std. Error 0.420685 0.138653 0.214719

t-Statistic 26.02720 26.93548 -5.923718

Wald Test: Equation: Untitled

Null Hypothesis Summary:

Restrictions are linear in coefficients.

WALD TEST MANUALLY (USING EVIEWS):

Unrestricted model: lnYi = 0 + 2 ln Ki + 3 ln Li + i

Variable C LOG(K) LOG(L) R-squared Adjusted R-squared

Std. Error 0.420685 0.138653 0.214719

t-Statistic 26.02720 26.93548 -5.923718

Prob. 0.0000 0.0000 0.0000 18.36495 0.432456

Mean dependent var S.D. dependent var

S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.025746 0.033143 120.2927 0.054050

Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

-4.426138 -4.314612 7310.653 0.000000

Restricted model: ln (Yi /Li) = 0 + 2 ln (Ki / Li) + i

0.959412 0.958617 0.069277 0.244764 67.30666

14.30042 0.340546 -2.464402 -2.390052 1205.542

F = {0.244764 0.033143 / 1} / {0.033143 / (53-3)}

The above example shows that F test is significant 1% level of significance.

Dependent Variable: SAV Method: Least Squares

Date: 08/21/13 Time: 14:08 Sample: 1960 2012 Included observations: 53

0.957821 0.956994 0.880589 39.54728 -67.44468 0.024164

29.38251 4.246259 2.620554 2.694904 1158.121 0.000000

Chow Breakpoint Test: 1980

F-statistic Log likelihood ratio

The above example shows that F test is significant 1% level of significance

MODEL DETERMINANTS OF LIFE EXPECTANCY IN PAKISTAN:

-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192

1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846

-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813

0.0002 0.0000 0.0000 0.0152 0.0000 0.2150

R-squared Adjusted R-squared

Mean dependent var S.D. dependent var

S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.295091 4.092694 -7.334900 0.196325

Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

0.503204 0.726256 3456.958 0.000000

CHAPTER 10 MULTICOLLINEARITY: WHAT HAPPENS REGRESSORS ARE CORRELATED? IF THE

DOING REGREESION USING EVIEWS FOR MULTICOLLINEARITY

1. HIGH R2 BUT FEW SIGNIFICANT t- RATIO:

MODEL: Yi = 1 + 2 X1i + 3 X2i + 4 X3i +5X4i + 6X5i +i

-5.519635 -1.20E-07 -1.61E-11 -0.215865 2.668870 -0.163192

1.350874 9.29E-09 2.45E-12 0.085649 0.070082 0.129846

-4.085972 -12.86161 -6.580235 -2.520339 38.08228 -1.256813

0.0002 0.0000 0.0000 0.0152 0.0000 0.2150

0.997288 0.997000 0.295091 4.092694 -7.334900

58.49076 5.387357 0.503204 0.726256 3456.958

HIGH PAIR WISE CORRELATION AMONG REGRESSORS:

The above correlation

matrix shows that X1 and X4 are highly correlated.

2. EXAMINATION OF PARTIAL CORRELATION:

Mean dependent var

Mean dependent var

Mean dependent var

MEAN(R) = 2N1N2 /N +1 = 1334/52+1 = 25.653 VARIANCE () = 2N1N2 (2N1N2 N) / N2 (N 1) = 13341282/270451