7. Prediction
Read Wooldridge (2013), Chapter 6.4
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
Outline: Read Section 6.4
• I. Mean Prediction
• II. Individual Prediction
• III. Predicting y from log(y)
• IV. Choose between y‐Model and log(y)‐Model
I. Mean II. Individual III. y^ from logy IV. R
2-log 2
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
I. Mean Prediction
• Predictions are useful
– But they are subject to sampling variations because they are obtained from the OLS estimators
• Goal: to find confidence intervals for a prediction.
1) Mean prediction – study a confidence interval around the OLS estimate of E(yx 1 , …, x k ), for instance, for the average hourly wage.
2) Individual prediction ‐ study a confidence interval of the hourly wage for a particular person.
I. Mean II. Individual III. y^ from logy IV. R
2-log 3
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Mean Prediction
Example: Find the predicted value of wage:
= ‐2.873 +.599educ + .022exper +.169tenure
s.e. (.729) (.0513) (.012) (.0216) t‐stat [‐3.94] [11.67] [1.85] [7.82]
n=526, R
2=.306422
• What is the predicted value when educ=16; exper=4; tenure=3?
= -2.873 +.599(16) + .022(4) +.169(3) = 7.3
• Claim: is the estimate of the expected value of wage given the particular value of explanatory variables. Thus, this value is called mean prediction.
Question: What is the s.e. of ?
Define =
I. Mean II. Individual III. y^ from logy IV. R
2-log 4
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Dependent Variable: WAGE Method: Least Squares Date: 05/22/03 Time: 11:39 Sample: 1 526
Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C -2.87274 0.728964 -3.940844 0.0001
EDUC 0.598965 0.051284 11.67948 0
EXPER 0.02234 0.012057 1.852849 0.0645
TENURE 0.169269 0.021645 7.820361 0
R-squared 0.306422 Mean dependent var 5.896103
Adjusted R-squared 0.302436 S.D. dependent var 3.693086
S.E. of regression 3.084476 Akaike info criterion 5.098216
Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F-statistic 76.87317
Durbin-Watson stat 1.791222 Prob(F-statistic) 0
Eviews: wage c educ exper tenure
I. Mean II. Individual III. y^ from logy IV. R
2-log 5
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
The use of se( ) or se( )
1) to test the statistical significance of the mean prediction.
2) to find the confidence interval of mean prediction.
• Given the estimated wage equation in y and x j form:
= + x 1 + x 2 + x 3
+ + + tenure
• Let educ = 16 x 1 = c 1
exper = 4 x 2 = c 2 tenure = 3 x 3 = c 3
• The parameter we would like to estimate is
0 = 0 + 1 c 1 + 2 c 2 + 3 c 3
I. Mean II. Individual III. y^ from logy IV. R
2-log 6
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Confidence Interval
• To find confidence interval, we need 1)
2) s.e( )
• A 95% confidence interval is
2*s.e( )
I. Mean II. Individual III. y^ from logy IV. R
2-log 7
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Trick to find the s.e.( )
• Trick : rewrite 0 in terms of 0 , c 1 , c 2 and c 3
0 = 0 + 1 c 1 + 2 c 2 + 3 c 3
0 = 0 ‐ 1 c 1 ‐ 2 c 2 ‐ 3 c 3
Substitute 0 = … into the y equation y= 0 + 1 x 1 + 2 x 2 + 3 x 3 + u
y = 0 + 1 (x 1 – c 1 ) + 2 (x 2 – c 2 ) + 3 (x 3 – c 3 ) + u
• Estimate to obtain s.e.( )
= + x 1 – c 1 ) + (x 2 – c 2 ) + (x 3 – c 3 )
• In Eviews, we find the estimate of s.e.( ) by running the regression of wage on (educ‐16) (exper‐4) (tenure‐3).
I. Mean II. Individual III. y^ from logy IV. R
2-log 8
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Dependent Variable: WAGE Method: Least Squares Date: 05/22/03 Time: 11:46 Sample: 1 526
Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C 7.30787 0.230378 31.72118 0
EDUC-16 0.598965 0.051284 11.67948 0
EXPER-4 0.02234 0.012057 1.852849 0.0645
TENURE-3 0.169269 0.021645 7.820361 0
R-squared 0.306422 Mean dependent var 5.896103
Adjusted R-squared 0.302436 S.D. dependent var 3.693086
S.E. of regression 3.084476 Akaike info criterion 5.098216
Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F-statistic 76.87317
Durbin-Watson stat 1.791222 Prob(F-statistic) 0
Eviews: wage c educ-16 exper-4 tenure-3
I. Mean II. Individual III. y^ from logy IV. R
2-log 9
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
t‐Stat and Confidence Interval for predicting the average value of hourly wage
1. Is the mean prediction statistically significant at the 1% significance level?
H
0:
0= 0
t = ( ‐
0)/se( )
= 7.308/0.230
= 31.72118 p‐value = 0
• Note that the predicted value, its standard error, t‐statistic, and p‐value are obtained from the intercept of the transformed equation.
I. Mean II. Individual III. y^ from logy IV. R
2-log 10
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Dependent Variable: WAGE Method: Least Squares Date: 05/22/03 Time: 11:46 Sample: 1 526 Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C 7.30787 0.230378 31.72118 0
EDUC-16 0.598965 0.051284 11.67948 0
EXPER-4 0.02234 0.012057 1.852849 0.0645
TENURE-3 0.169269 0.021645 7.820361 0
R-squared 0.306422 Mean dependent var 5.896103
Adjusted R-squared 0.302436 S.D. dependent var 3.693086 S.E. of regression 3.084476 Akaike info criterion 5.098216 Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F-statistic 76.87317
Durbin-Watson stat 1.791222 Prob(F-statistic) 0
Interpret: 2 s.e.( )
2. What is the 95% CI for predicting the average value of hourly wage,
0?
2*se( )
= 7.31 2(0.2304)
= (6.85, 7.77) dollars per hour
• INTERPRETATION: The predicted mean wage ranges from 6.85 to 7.7 dollars per hour with the 95%
confidence.
I. Mean II. Individual III. y^ from logy IV. R
2-log 11
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
Dependent Variable: WAGE Method: Least Squares Date: 05/22/03 Time: 11:46 Sample: 1 526 Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C 7.30787 0.230378 31.72118 0
EDUC-16 0.598965 0.051284 11.67948 0
EXPER-4 0.02234 0.012057 1.852849 0.0645
TENURE-3 0.169269 0.021645 7.820361 0
R-squared 0.306422 Mean dependent var 5.896103
Adjusted R-squared 0.302436 S.D. dependent var 3.693086 S.E. of regression 3.084476 Akaike info criterion 5.098216 Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F-statistic 76.87317
Durbin-Watson stat 1.791222 Prob(F-statistic) 0
Compare Results from Eviews
Original Equation:
= ‐2.873 +.599educ + .022exper +.169tenure s.e. (.729) (.0513) (.012) (1.85) t‐stat [‐3.94] [11.67] [1.85] [7.82]
n=526, R
2=.306422 Transformed Equation to Find s.e.( )
= 7.31 +.599(educ‐16) + .022(exper‐4) +.169(tenure‐3) s.e. (.230) (.0513) (.012) (1.85) t‐stat [31.72] [11.67] [1.85] [7.82]
n=526, R
2=.306422
I. Mean II. Individual III. y^ from logy IV. R
2-log 12
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat I. Mean Prediction
II. Prediction Interval
• Mean Prediction – confidence interval for the average hourly wage.
Individual Prediction –confidence interval of the hourly wage for a particular person, This is called prediction interval.
• What is the unknown value of y (wage) when x
1=x
10, x
2=x
20and x
3=x
30?
y
0=
0+
1x
10+
2x
20+
3x
30+ u
0• What is the predicted value of y when x
1=x
10, x
2=x
20and x
3=x
30?
0
= + x
10+ x
20+ x
30I. Mean II. Individual III. y^ from logy IV. R
2-log 13
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
Find a 95% CI for an unknown y
• The prediction error in using y
0^ to predict y
0is
̂
0= y
0‐
0=
0+
1x
10+
2x
20+
3x
30+ u
0‐
0• Mean and Variance of ̂
0E( ̂
0) = 0
Var( ̂
0) = Var(
0) +
2I. Mean II. Individual III. y^ from logy IV. R
2-log 14
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
• What is the value of s.e.( ̂
0) found in Eviews?
s.e.( ̂
0) = [s.e.(
0)
2 + 2]
1/2is S.E. of regression (=3.084475)
s.e.(
0) from the previous section (=0.2304)
s.e.( ̂
0) = 3.09
Dependent Variable: WAGE Method: Least Squares Date: 05/22/03 Time: 11:39 Sample: 1 526
Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C -2.87274 0.728964 -3.940844 0.0001
EDUC 0.598965 0.051284 11.67948 0
EXPER 0.02234 0.012057 1.852849 0.0645
TENURE 0.169269 0.021645 7.820361 0
R-squared 0.306422 Mean dependent var 5.896103
Adjusted R-squared 0.302436 S.D. dependent var 3.693086
S.E. of regression 3.084476 Akaike info criterion 5.098216
Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F-statistic 76.87317
Durbin-Watson stat 1.791222 Prob(F-statistic) 0
Eviews: wage c educ exper tenure
I. Mean II. Individual III. y^ from logy IV. R
2-log 15
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
16 Dependent Variable: WAGE
Method: Least Squares Date: 05/22/03 Time: 11:39 Sample: 1 526 Included observations: 526
Variable Coefficient Std. Error t-Statistic Prob.
C -2.87274 0.728964 -3.940844 0.0001
EDUC 0.598965 0.051284 11.67948 0
EXPER 0.02234 0.012057 1.852849 0.0645
TENURE 0.169269 0.021645 7.820361 0
R‐squared 0.306422 Mean dependent var 5.896103 Adjusted R‐squared 0.302436 S.D. dependent var 3.693086 S.E. of regression 3.084476 Akaike info criterion 5.098216 Sum squared resid 4966.303 Schwarz criterion 5.130652
Log likelihood -1336.83 F‐statistic 76.87317
Durbin‐Watson stat 1.791222 Prob(F‐statistic) 0
t‐Statistic and CI
• Note that ̂
0is normally distributed with mean 0 and variance, Var( ̂
0):
̂
0~ Normal[0,Var( ̂
0)]
• The t‐statistic is t=
0/s.e.(
0),
which has a t distribution with n‐k‐1 DFs.
• A 95% confidence interval is P(‐t
0.025< t < t
0.025) = .95
Thus , a 95% confidence interval for unknown y
0is
0 2*s.e.(
0).
I. Mean II. Individual III. y^ from logy IV. R
2-log 17
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
Example: Individual Prediction of wage
• Find a 95% Confidence interval for an unknown hourly wage when educ
0=x
10=16, exper
0=x
20=4 and tenure
0=x
30=3
• The prediction of wage
0is
0
= y
0= 7.31 se( ̂
0) = 3.09
• A confidence interval of y
0is
0
2*se( ̂
0) = 7.31 2*3.09 (1.24, 13.36) dollars per hour
I. Mean II. Individual III. y^ from logy IV. R
2-log 18
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
Mean Prediction versus Individual Prediction Compared
• Mean Prediction
– Find a confidence interval around the OLS estimate of E(yx
1,…,x
k) for any values of regressors.
One Source of variation: the variance of sampling error.
• Individual Prediction
– Find a confidence interval for a particular unit (individual, family, firm, and so on).
– y
0could be a person or firm not in our original sample.
Two sources of variations:
(1) the variance of sampling error and
(2) the variance of unobserved error (in the population)
I. Mean II. Individual III. y^ from logy IV. R
2-log 19
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat II. Prediction Interval
III. Predicting y from log(y)
• Suppose we have a model
log(salary) = 0 + 1 log(sales)+ 2 log(mktval)+ 3 ceoten+u
• General form:
log(y) = 0 + 1 x 1 + 2 x 2 + 3 x 3 + u x 1 = log(sales)
x 2 = log(mktval) x 3 = ceoten
• Estimate
log( ) = + x 1 + x 2 + x 3
I. Mean II. Individual III. y^ from logy IV. R
2-log 20
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Easy way but inexact! for predicting y
= exp[log( )]
It will underestimate the expected value of y.
The correct method to find predicted y is
= exp[log( )]
Motivation why we need to find
With assumptions MLR1-MLR6, it can be shown that y = exp(u)*exp(
0+
1x
1+
2x
2+
3x
3) E(y|x) = exp(
2/2)*exp(
0+
1x
1+
2x
2+
3x
3)
The sample counterpart is
= exp(
2/2)*exp(log( ))
I. Mean II. Individual III. y^ from logy IV. R
2-log 21
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Given = exp[log( )]
We can find the coefficient by regressing y i on exp[log( )].
log( ) = 4.504 + .163log(sales) +.109log(mktval) +.0117ceoten log( ) = 7.013
= 1,110.983 (inexact value = exp[log( ))]
The correct method to find predicted y is
= exp[log( )]
= exp[7.01] How to find ?
I. Mean II. Individual III. y^ from logy IV. R
2-log 22
Example: Prediction of CEO when sales=5,000, mktal=10,000 and ceoten=10
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Eviews: Steps to find from log( ) where
= exp[log( )]
(1) Given values of x 1 0 , x 2 0 , x 3 0 , obtain log( ).
(2) Find the fitted values of log(y) from the estimation equation
log(y) = 0 + 1 x 1 + 2 x 2 + 3 x 3 + u to obtain exp[log( )].
(3) Find from regression with no intercept.
regress y on exp[log( )]
(4) After is known, we can find the predicted value of y.
I. Mean II. Individual III. y^ from logy IV. R
2-log 23
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Example
• Step (1): Estimate the model to obtain the predicted value of log(salary) or lsalary when
sales = 5,000 log(sales) = 8.517193 mktval = 10,000 log(mktval) = 9.21034
ceoten = 10 ceoten = 10
= 4.504 +.163lsales +.109lmktval + .0117ceoten n =177 R
2= 00.318
log( ) = lsalaryhat = 7.013
= 1,110.983 (inexact value = exp[log( )])
I. Mean II. Individual III. y^ from logy IV. R
2-log 24
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Dependent Variable: LSALARY Method: Least Squares Date: 07/05/03 Time: 20:42 Sample: 1 177
Included observations: 177
Variable Coefficient Std. Error t-Statistic Prob.
C 4.503795 0.257234 17.50852 0
LSALES 0.162854 0.039242 4.149993 0.0001
LMKTVAL 0.109243 0.049595 2.202717 0.0289
CEOTEN 0.011705 0.005326 2.19776 0.0293
R-squared 0.318151 Mean dependent var 6.582848
Adjusted R-squared 0.306327 S.D. dependent var 0.606059
S.E. of regression 0.504769 Akaike info criterion 1.492908
Sum squared resid 44.07898 Schwarz criterion 1.564685
Log likelihood -128.122 F-statistic 26.90726
Durbin-Watson stat 2.043631 Prob(F-statistic) 0
Step 1: lsalary c lsales lmktval ceoten
Proc/Make Residuals Series … use the default name, “resid01”
I. Mean II. Individual III. y^ from logy IV. R
2-log 25
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
obs Actual Fitted Residual Residual Plot
1 7.057037 7.047401 0.009636 | . * . | 2 6.39693 6.305271 0.091659 | . * . | 3 5.937536 6.139374 -0.20184 | .*| . |
.. .. ..
174 5.220356 6.288329 -1.06797 | * . | . | 175 5.958425 6.317993 -0.35957 | * | . | 176 7.705262 6.317698 1.387564 | . | . * | 177 6.098074 6.136342 -0.03827 | . * . |
Step 2: obtained the fitted values in Eviews by generating a series: lsalaryhat=lsalary-resid01
I. Mean II. Individual III. y^ from logy IV. R
2-log 26
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Step (3): Find by regressing y on exp[log( ))
In Eviews, regress salary on exp(lsalaryhat with no intercept.
Dependent Variable: SALARY Method: Least Squares Date: 07/05/03 Time: 20:57 Sample: 1 177
Included observations: 177
Variable Coefficient Std. Error t-Statistic Prob.
EXP(LSALARYHAT) 1.116857 0.047095 23.7148 0
R-squared 0.241123 Mean dependent var 865.8644
Adjusted R-squared 0.241123 S.D. dependent var 587.5893
S.E. of regression 511.8699 Akaike info criterion 15.31965
Sum squared resid 46113901 Schwarz criterion 15.3376
Log likelihood -1354.79 Durbin-Watson stat 2.08316
I. Mean II. Individual III. y^ from logy IV. R
2-log 27
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
• Step 4: Find the predicted value of y
log( ) = lsalaryhat = 7.013
exp(log( )) = exp(lsalaryhat) = 1,110,983
= exp[log( )]
salaryhat = exp[lsalaryhat]
salaryhat = 1.117[exp(7.013)]
= 1.117[1,110.982]
= 1,240.967
In dollars, the predicted salary is $1,240,967.
I. Mean II. Individual III. y^ from logy IV. R
2-log 28
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Which model is preferred?
• Example: CEO salary of 177 firms
salary 1990 compensation, $1000s sales 1990 firm sales, millions
mktval market value, end 1990, millions.
ceoten years as CEO with company
log( ) = 4.504 + .163log(sales) +.109log(mktval) + .0117ceoten t‐stat [17.51] [4.15] [2.20] [2.20]
n=177, R
2=0.31815, R
2‐bar=.306327
= 613.43 +.019sales + .0234mktval + 12.703ceoten t‐stat [9.40] [1.89] [2.47] [2.26]
n=177, R
2= 0.201274, R
2‐bar=.187424
I. Mean II. Individual III. y^ from logy IV. R
2-log 29
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
IV. Comparing R 2 for dependent variables: y and log(y)
• Comparing two models:
log( ) = 4.504 +.163log(sales) +.109log(mktval)+.117log*(ceoten)
= 613.43 +.019sales + .0234mktval + 12.703ceoten
• log( ) = 4.504 +.163log(sales) +.109log(mktval)+.117log*(ceoten)
Trick : find the R 2 – equivalence from the log (y) equation (find salary variation)
R 2 = [corr(y i , i )] 2 is the square of the sample correlation between y i and
I. Mean II. Individual III. y^ from logy IV. R
2-log 30
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat IV. Comparing R2 for dependent variables: y and log(y)
R 2 for dependent variables
• R 2 = [corr(y i , i )] 2 y i = salary
i = fitted value of salary from log( ) equation (called salaryhat).
• In log(salary) equation, we can find
= exp[log( )]
• In Eviews, generating a new series, salaryhat = 1.117exp(lsalaryhat)
I. Mean II. Individual III. y^ from logy IV. R
2-log 31
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat IV. Comparing R2 for dependent variables: y and log(y)
Generating new series,
salaryhat =1.117exp(lsalaryhat)
Open two series – salary and salaryhat as “group”
obs SALARY SALARYHAT
1 1161 1284.401
2 600 611.5016
3 379 518.0234
.. ..
174 185 601.2289
175 387 619.3311
176 2220 619.1484
177 445 516.4551
I. Mean II. Individual III. y^ from logy IV. R
2-log 32
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat IV. Comparing R2 for dependent variables: y and log(y)
In “Group” window, Choose View/Correlation
Eviews: find the correlation between salaryhat and salary and R
2. corr(
i, y
i) = 0.493
R
2= [corr(
i, y
i)]
2= 0.243
SALARY SALARYHAT
SALARY 1 0.493032
SALARYHAT 0.493032 1
I. Mean II. Individual III. y^ from logy IV. R
2-log 33
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat IV. Comparing R2 for dependent variables: y and log(y)
R 2 ‐equivalence = 0.243
Log dependent variable:
log( ) = 4.504 + .163lsales + .109lmktval + .117ceoten R
log2=0.31815: variation in log(salary)
R
2‐equivalence = (0.49302)
2= 0.243
: variation in salary!
Level dependent variable:
= 613.43 +.019sales + .0234mktval + 12.703ceoten R
level2= 0.201274: variation in salary
Which model do you prefer?
I. Mean II. Individual III. y^ from logy IV. R
2-log 34
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat IV. Comparing R2 for dependent variables: y and log(y)
Dependent Variable: LSALARY Method: Least Squares Date: 07/05/03 Time: 20:42 Sample: 1 177
Included observations: 177
Variable Coefficient Std. Error t-Statistic Prob.
C 4.503795 0.257234 17.50852 0
LSALES 0.162854 0.039242 4.149993 0.0001
LMKTVAL 0.109243 0.049595 2.202717 0.0289
CEOTEN 0.011705 0.005326 2.19776 0.0293
R-squared 0.318151 Mean dependent var 6.582848
Adjusted R-squared 0.306327 S.D. dependent var 0.606059
S.E. of regression 0.504769 Akaike info criterion 1.492908
Sum squared resid 44.07898 Schwarz criterion 1.564685
Log likelihood -128.122 F-statistic 26.90726
Durbin-Watson stat 2.043631 Prob(F-statistic) 0
Eviews: lsalary c lsales lmktval ceoten
I. Mean II. Individual III. y^ from logy IV. R
2-log 35
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Eviews: salary sales mktval ceoten c
Dependent Variable: SALARY Method: Least Squares Date: 07/05/03 Time: 21:06 Sample: 1 177
Included observations: 177
Variable Coefficient Std. Error t-Statistic Prob.
C 613.4361 65.23685 9.403214 0
SALES 0.019019 0.010056 1.89129 0.0603
MKTVAL 0.0234 0.009483 2.467719 0.0146
CEOTEN 12.70337 5.618052 2.261169 0.025
R-squared 0.201274 Mean dependent var 865.8644
Adjusted R-squared 0.187424 S.D. dependent var 587.5893 S.E. of regression 529.6707 Akaike info criterion 15.40473
Sum squared resid 48535332 Schwarz criterion 15.47651
Log likelihood -1359.32 F-statistic 14.53168
Durbin-Watson stat 2.166043 Prob(F-statistic) 0
I. Mean II. Individual III. y^ from logy IV. R
2-log 36
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat III. Predicting y from log(y)
Recap of Prediction
• Mean Prediction
• Individual Prediction
• Predicting y from log(y)
• Choose between y‐Model and log(y)‐Model
I. Mean II. Individual III. y^ from logy IV. R
2-log 37
7. Prediction . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat