ECON 644 Class 6

Class 6
Multiple Regression
Inference II

Readings:
Wooldridge (2025) Chapter 4
 ECON 644 Class 6

Part I

(1) Hypothesis Type 2

(2) Does a t-test Still Work?

(3) Direct Method vs. Indirect Method

 ECON 644 Class 6

Types of Hypotheses

Type 1 One Coefficient: 1 .

Type 2 One Linear Combination of Coefficients: 1  2 .

Type 3 Multiple Linear Combinations of Coefficients: 1  2 ,

31  22 , etc.
 ECON 644 Class 6

(1) Hypotheses Type 2

H 0 : c0  0  c11  c2  2  ...  ck  k  c

Note: The expression on the left c0  0  c11  c2  2  ...  ck  k is

called a linear combination of the  s. In words, a sum of
constant multiples of the  s.

Example: Is the return to CEO tenure at least two times greater

than the return to age?

H 0 : 1  22
H1 : 1  22
 ECON 644 Class 6

(2) Does a t-test Still Work?

We could think of the linear combination as a new population

feature, call it  :

  c0  0  c11  c2  2  ...  ck  k  c

Then, the null hypothesis becomes:

H0 :  0

Question: This looks like a Hypothesis Type 1 (Test of

Significance)! But does the t-test still work?
 ECON 644 Class 6

Does a t-test Still Work?

To set up a t-test, we need three parts:

 an unbiased estimator for  : ˆ

 the standard error of ˆ : se(ˆ )

 the probability distribution of ˆ .

 ECON 644 Class 6

Does a t-test Still Work?

 The estimator ˆ is easy to get, because (as is shown in

Problem 3.6) if ˆ0 , ˆ1 , ˆ2 ,..., ˆk are unbiased estimators of
 0 , 1 ,  2 ,...,  k then
c0 ˆ0  c1ˆ1  c2 ˆ2  ...  ck ˆk  c
is an unbiased estimator of
c0  0  c11  c2  2  ...  ck  k  c .

 It is somewhat more difficult to get the standard error se(ˆ ),

because it depends on the correlations between
ˆ0 , ˆ1 , ˆ2 ,..., ˆk , and these different estimators are generally
not independent.

ˆ
 It can be shown that, under MLR.1-MLR.6, ~ tnk 1 . This
ˆ
se( )
is our test statistic.
 ECON 644 Class 6

F-test vs. t-test

Note: The textbook does these tests with a tn k 1 statistic.

Stata does these tests with an F (1,n  k  1) statistic.

Remember, though, that F (1,n  k  1)  tn k 12 , so it’s really easy to

switch between the two methods.

For large samples, statistical significance is attained when:

 t-stat > 1.96 (or approximately 2).
 F-stat > 3.84 (or approximately 4).
 ECON 644 Class 6

Tenure vs. Age

. regress salary ceoten age grad mktval

Source | SS df MS Number of obs = 177

-------------+------------------------------ F( 4, 172) = 9.87
Model | 11348587.4 4 2837146.84 Prob > F = 0.0000
Residual | 49417377.4 172 287310.334 R-squared = 0.1868
-------------+------------------------------ Adj R-squared = 0.1678
Total | 60765964.7 176 345261.163 Root MSE = 536.01

------------------------------------------------------------------------------
salary | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ceoten | 10.55413 6.019706 1.75 0.081 -1.327879 22.43614
age | 1.649817 5.162111 0.32 0.750 -8.539427 11.83906
grad | -43.43346 82.2966 -0.53 0.598 -205.8748 119.0079
mktval | .0371649 .0063706 5.83 0.000 .0245903 .0497395
_cons | 578.0711 288.3528 2.00 0.047 8.905383 1147.237
------------------------------------------------------------------------------

Note: In this sample:

Return to tenure $10,554.
Return to age $1,649.
So, in the sample, the return to tenure is at least twice the return
to age.
 ECON 644 Class 6

Tenure vs. Age

Question: But is that also true in the population?

. test (ceoten = 2*age)

( 1) ceoten - 2*age = 0

F( 1, 172) = 0.29
Prob > F = 0.5934

Answer: p-value not below 0.05. Thus, there is not enough

evidence that, in the population, tenure is rewarded more than
twice as much as age.
 ECON 644 Class 6

Another Example:
Return to Education Depends on Type of School?

Question: Does a university education increase wage more than

a community college education?

A possible model for testing for this is:

where:
wage  yearly wage, in current dollars
jc  time enrolled at a junior college, in years
univ  time enrolled at a university, in years
exper  number of months in the workforce
 ECON 644 Class 6

Junior College vs. University

. regress lwage jc univ exper

Source | SS df MS Number of obs = 6763

-------------+------------------------------ F( 3, 6759) = 644.53
Model | 357.752575 3 119.250858 Prob > F = 0.0000
Residual | 1250.54352 6759 .185019014 R-squared = 0.2224
-------------+------------------------------ Adj R-squared = 0.2221
Total | 1608.29609 6762 .237843255 Root MSE = .43014

------------------------------------------------------------------------------
lwage | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
jc | .0666967 .0068288 9.77 0.000 .0533101 .0800833
univ | .0768762 .0023087 33.30 0.000 .0723504 .0814021
exper | .0049442 .0001575 31.40 0.000 .0046355 .0052529
_cons | 1.472326 .0210602 69.91 0.000 1.431041 1.51361
------------------------------------------------------------------------------

Note: In this sample:

Return to comm college education: 6.67%
Return to university education: 7.69%.

Question 1: Are 6.67% and 7.69% statistically significant?

Question 2: Is the difference between them (1.02pp) statistically
significant?
 ECON 644 Class 6

Junior College vs. University

Question 1: Are 6.67% and 7.69% statistically significant?

. test (jc = 0) . test (univ = 0)

( 1) jc = 0 ( 1) univ = 0

F( 1, 6759) = 95.39 F( 1, 6759) = 1108.76

Prob > F = 0.0000 Prob > F = 0.0000
Answer: Yes, p-values below 0.05. You can also see the p-
values in the regression output above. (These are Hyp Type 1.)

Question 2: Is the difference between them (1.02pp) statistically

significant?
. test (jc = univ)

( 1) jc - univ = 0

F( 1, 6759) = 2.15
Prob > F = 0.1422
Answer: No, p-value not below 0.05. (This is Hyp Type 2.)
 ECON 644 Class 6

Why Does a t-test Still Work?

The claim that the return to community college ( 1) is smaller

than the return to university (  2 ) can be tested as:

H 0 : 1   2
H1 : 1   2
 ECON 644 Class 6

A t-test Still Works

1) Create a new parameter:

  1   2

2) Restate the hypothesis:

H 0 :   0 against H1 :   0

3) Run a usual t-test:

 An unbiased estimator of   1   2 is
ˆ  ˆ1  ˆ2 .
 The standard error of ˆ can be found in two ways:
1) directly: se(ˆ )=se( ˆ1  ˆ2 )
2) indirectly: by setting up a model in which  is a slope
parameter.
ˆ
 The test statistic is t  .
ˆ
se( )
 ECON 644 Class 6

(3) Direct Method to Find se(ˆ )

Since ˆ  ˆ1  ˆ2 it follows that:

se(ˆ )=se( ˆ1  ˆ2 )= SmplVar( ˆ1  ˆ2 ) =
= SmplVar( ˆ1 )  SmplVar( ˆ2 )  2SmplCov( ˆ1 , ˆ2 )
where:
2
ˆ ˆ
SmplVar( 1 )  se( 1 ) 
2
ˆ ˆ
SmplVar(  2 )  se(  2 ) 

Note: The standard regression output will report se( ˆ1 ) and
se( ˆ2 ) , but not SmplCov( ˆ1 , ˆ2 ) . This sample covariance must
be computed separately.
 ECON 644 Class 6

Indirect Method to Find se(ˆ )

(1) Original model:

(2) Substitute  for 1 using the fact that 1     2 .

(3) Modified model:

Note: But this is exactly what we wanted: to have  itself as a

slope coefficient in a regression model.

(4) Estimate this modified model by OLS and test ˆ for statistical
significance.
 ECON 644 Class 6

Indirect Method to Find se(ˆ )

OLS estimation of the modified model:

Question: What is ˆ ? What is se(ˆ ) ?

Answer: ˆ  .0102 and se(ˆ)  .0069 .

Question: Is ˆ statistically significant?

0.0102
Answer: No, the t-stat is   1.478, whose abs val  1.96 .
0.0069

Conclusion: No evidence for a larger return from university.

 ECON 644 Class 6

Conclusion

Conclusion: Good news. A t -test works both for:

 Hypothesis Type 1
 Hypothesis Type 2

Preview: It will not, however, work for Hypothesis Type 3. There,

we need a true F-test.
 ECON 644 Class 6

Part II

(1) Hypothesis Type 3

joint significance
overall model significance
multiple linear combinations

(2) The Restricted/Unrestricted Approach

(3) Examples
Does talent pay in baseball?
Is the stock market "efficient"?
Are house prices "rational"?
 ECON 644 Class 6

(1) Hypothesis Type 3

Examples

salary   0  1ceoten   2 age   3 grad   4 mktval  u

Are the tenure and age coefficients jointly significant?

H 0 : 1   2  0

Is the model overall significant?

H 0 : 1   2  3   4  0

Is the effect of tenure more than double the effect of age, and
their cumulative effect larger than the effect of market value?

H 0 : 1  2  2 and 1   2   4
 ECON 644 Class 6

What’s the Common Thread in These Examples?

They can all be written as a list of claims about linear

combinations of parameters:

H 0 : LC 1  c1 and LC 2  c2 and ... LC q  cq

H 1 : at least one of these is not true.

Question: Could we test this type of hypothesis using a standard

t-test?
Answer: The t-test works when we have only one LC. For
multiple LCs we need a different approach.
 ECON 644 Class 6

(2) The Restricted/Unrestricted Approach

Unrestricted Regression Model

+
LCs

=
Restricted Regression Model

The Logic of the Test:

1) Compare the fit of the restricted model (r ) to the fit of
the original unrestricted model (ur ).
2) If restrictions reduced the fit, it means they pushed the
model further away from the data, so they are not
consistent with the data.
 fit statistically smaller  reject restrictions.
 fit not statistically smaller  can’t reject restrictions.
 ECON 644 Class 6

Testing H 0 : (q linear restrictions)

H 1 : ( at least one restriction not true)
1) Set up the null and alternative hypotheses.
 SSRr  SSRur  / q
2) An appropriate test statistic is which can be
SSRur /( n  k  1)
shown to have an Fq ,n  k 1 distribution under the null hypothesis.
3) Adopt a reasonable confidence level (say, 95%) and find the
threshold Fq ,n  k 1 (0.05) in the Fq ,n  k 1 distribution beyond which
the probability of getting a draw from Fq ,n  k 1 is 5% . The rejection
region is Fq ,n  k 1  Fq ,n  k 1 (0.05) (why? see next slide).
4) Reject H 0 if the test statistic in the rejection region. Otherwise,
don't reject H 0 .
---------------------------------------------------------------------------------------

3) Compute the p  value of the test: P Fq ,n  k 1  test  stat 
4) Reject H 0 if the p  value is low enough.
 ECON 644 Class 6

Note: The test statistic

 SSRr  SSRur  / q
is always positive (b/c
SSRur /( n  k  1)
the restricted model always has poorer fit, SSRr  SSRur ).
Therefore the F test is always one-sided, even if the null and
alternative hypotheses look two-sided.
Reject null if difference in fit is significantly different from zero.
. display invF(3,60,.95)
2.7580783
 ECON 644 Class 6

Two Formulas for the Test Statistic

SSR version:
F
 SSRr  SSRur  / q
SSRur / ( n  k  1)

R-sq version:
( Rur 2  Rr 2 ) / q
F
1  Rur 2  / (n  k  1)

SSR
Remember: R 2  1  .
SST
 ECON 644 Class 6

(3) Example 1: Joint Significance Test

A baseball player's pay may depend both on:

 "effort" (like years in the league, games per year)
 "talent" (stats like career batting average, home runs per
year, runs batted in per year)

Research Question: Given equal effort, does "talent" pay?

H 0 : 3   4  5  0
H 1 : at least one is nonzero.

Note: This is a test of “joint” significance for  3 ,  4 ,  5 .

 ECON 644 Class 6

Does Talent Pay in Baseball?

1) The null and alternative hypotheses:
H 0 : 3   4  5  0
H 1 : at least one is nonzero.
2) Estimate:
 unrestricted model:

 restricted model:

Test statistic: F 
 SSRr  SSRur  / q (198.311  183.186) / 3
  9.55
SSRur / ( n  k  1) 183.186 / (353  5  1)
3) Threshold F3,347 (0.05)  2.6 .
4) Test statistic beyond threshold -> Reject H 0 .

Conclusion: "Talent" does pay in baseball.

 ECON 644 Class 6

Test for Joint Significance in Stata®

. regress lsalary years gamesyr bavg hrunsyr rbisyr

Source | SS df MS Number of obs = 353

-------------+------------------------------ F( 5, 347) = 117.06
Model | 308.989208 5 61.7978416 Prob > F = 0.0000
Residual | 183.186327 347 .527914487 R-squared = 0.6278
-------------+------------------------------ Adj R-squared = 0.6224
Total | 492.175535 352 1.39822595 Root MSE = .72658

------------------------------------------------------------------------------
lsalary | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
years | .0688626 .0121145 5.68 0.000 .0450355 .0926898
gamesyr | .0125521 .0026468 4.74 0.000 .0073464 .0177578
bavg | .0009786 .0011035 0.89 0.376 -.0011918 .003149
hrunsyr | .0144295 .016057 0.90 0.369 -.0171518 .0460107
rbisyr | .0107657 .007175 1.50 0.134 -.0033462 .0248776
_cons | 11.19242 .2888229 38.75 0.000 10.62435 11.76048
------------------------------------------------------------------------------

Statistics > Postestimation > Tests > Test linear hypotheses > Create … Test type (Coefficients are 0)

. test (bavg hrunsyr rbisyr)

( 1) bavg = 0
( 2) hrunsyr = 0
( 3) rbisyr = 0

F( 3, 347) = 9.55
Prob > F = 0.0000
 ECON 644 Class 6

Joint vs. Individual Significance

Interesting Observation:
 none of the three "talent" variables is individually
significant (p-values above 0.05)
 however, as a group, they are jointly significant.

Question: How can this be?

Answer: High Collinearity.
. corr bavg hrunsyr rbisyr
(obs=353)
| bavg hrunsyr rbisyr
-------------+---------------------------
bavg | 1.0000
hrunsyr | 0.1906 1.0000
rbisyr | 0.3291 0.8907 1.0000

Home runs per year ( hrunsyr ) and runs batted in per year
( rbisyr ) are highly correlated. This makes their OLS std. errors
very high -> their individual effects are imprecisely estimated.
 ECON 644 Class 6

Example 2: Overall Model Significance

Research Question: Is the Stock Market efficient?

return = return on a company’s stock in 1994

dkr, eps, netinc, salary = company financial info from 1990.

return   0  1dkr   2 eps   3 netinc   4 salary  u

H 0 : 1   2   3   4  0
H 1 : at least one is nonzero.

Note: This is a test of “joint” significance for all slopes

1 ,  2 ,  3 ,  4 .
 ECON 644 Class 6

Nobel Prize in Economics

2013 Winners
 ECON 644 Class 6

Nobel Prize in Economics

2013 Winners

Eugene F. FAMA Lars Peter HANSEN Robert J. SHILLER

U. Chicago, Booth School of Business U. Chicago, Economics Department Yale U., Economics Department

The Stock Market The Stock Market

is Efficient. is Not Efficient.
 ECON 644 Class 6

Is the Stock Market Efficient?

1) Null and alternative hypotheses:

H 0 : 1   2   3   4  0
H 1 : at least one is nonzero.
2) The restricted model is:

which, by the way, has no explanatory variables. The SSR of this

(trivial) model is actually equal to SST , so the test statistic is:
Rur 2 / q 0.0395 / 4
F   1.41
1  Rur  / (n  k  1) 1  0.0395 / 142  4  1
2

3) Threshold: F4,14241 (0.05)  2.37

4) Test stat 1.41 not beyond threshold -> Can’t reject H 0 .

Conclusion: We can’t conclude at 95% confidence whether the

stock market is efficient or not. Data is consistent with efficiency.
 ECON 644 Class 6

Test for Model Significance in Stata®

. regress return dkr eps netinc salary

Source | SS df MS Number of obs = 142

-------------+------------------------------ F( 4, 137) = 1.41
Model | 8649.26028 4 2162.31507 Prob > F = 0.2347
Residual | 210446.917 137 1536.10888 R-squared = 0.0395
-------------+------------------------------ Adj R-squared = 0.0114
Total | 219096.178 141 1553.8736 Root MSE = 39.193

------------------------------------------------------------------------------
return | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
dkr | .3205444 .2009106 1.60 0.113 -.0767426 .7178314
eps | .0426986 .0781384 0.55 0.586 -.1118147 .1972119
netinc | -.0051086 .0046748 -1.09 0.276 -.0143526 .0041354
salary | .0034993 .0021935 1.60 0.113 -.0008382 .0078369
_cons | -14.37022 6.893616 -2.08 0.039 -28.00187 -.7385647
------------------------------------------------------------------------------

H 0 : all slope coefficients are zero

H 1 : at least one is nonzero.
 ECON 644 Class 6

Joint vs. Individual Significance

Observation:
 none of the four "past info" variables have an
individually significant effect on current stock
returns (p-values above 0.05)
 as a group, they are not jointly significant either.
 ECON 644 Class 6

Example 3: Several Linear Combinations

Research Question: Are house prices “rational”?

price = house market price

assess = house assessed value
lotsize, sqrft, bdrms = house features

H 0 : 1  1,  2   3   4  0
H 1 : at least one linear combination is not true.

Note: This is a test of multiple (four) linear combinations.

 ECON 644 Class 6

Are House Prices "Rational"?

1) The null and alternative hypotheses are:

H 0 : 1  1,  2  0,  3  0,  4  0
H 1 : at least one linear combination is not true.
2) Estimate:
 the unrestricted model: SSRur  1.822
 the restricted model (please check that this is indeed the
restricted model): SSRr  1.880

Test statistic: F 
 SSRr  SSRur  / q (1.880  1.822) / 4
  0.661
SSRur / ( n  k  1) 1.822 / (88  4  1)
3) Threshold F4,83 (0.05)  2.50 .
4) Test statistic not beyond threshold, can’t reject H 0 .

Conclusion: We can’t conclude at 95% confidence whether

home values are rational. Data consistent with rationality.
 ECON 644 Class 6

Test for Multiple LCs in Stata®

. regress lprice lassess llotsize lsqrft bdrms

Source | SS df MS Number of obs = 88

-------------+------------------------------ F( 4, 83) = 70.58
Model | 6.19607473 4 1.54901868 Prob > F = 0.0000
Residual | 1.82152879 83 .02194613 R-squared = 0.7728
-------------+------------------------------ Adj R-squared = 0.7619
Total | 8.01760352 87 .092156362 Root MSE = .14814

------------------------------------------------------------------------------
lprice | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lassess | 1.043065 .151446 6.89 0.000 .7418453 1.344285
llotsize | .0074379 .0385615 0.19 0.848 -.0692593 .0841352
lsqrft | -.1032384 .1384305 -0.75 0.458 -.378571 .1720942
bdrms | .0338392 .0220983 1.53 0.129 -.0101135 .0777918
_cons | .263743 .5696647 0.46 0.645 -.8692972 1.396783
------------------------------------------------------------------------------

Statistics > Postestimation > Tests > Test linear hypotheses > Create … Test type (Linear expressions
are equal)
. test (lassess-1=llotsize=lsqrft=bdrms=0)

( 1) lassess - llotsize = 1
( 2) lassess - lsqrft = 1
( 3) lassess - bdrms = 1
( 4) lassess = 1

F( 4, 83) = 0.67
Prob > F = 0.6162
 ECON 644 Class 6

Summing-Up: Inference

Remember: All these testing procedures are valid only under

MLR.1-MLR.6. The key assumptions among those are:
 model is correctly identified (MLR.4)
 the homoskedasticity assumption (MLR.5).

We learnt how to perform three types of hypotheses tests:

Type 1 single coefficient use a t - test

Type 2 single linear combination use a t - test

(directly or indirectly)
Type 3 multiple linear combinations use an F - test
 ECON 644 Class 6

Part III

(1) Hypotheses for CEO Salary

(2) Stata Scalars

(3) Unpacking the t-stat and F-stat Calculations

 ECON 644 Class 6

(1) Hypotheses for CEO Salary

salary   0  1ceoten   2 age   3 grad   4 mktval  u

Are the tenure and age coefficients jointly significant?

H 0 : 1   2  0

Is the model overall significant?

H 0 : 1   2  3   4  0

Is the effect of tenure more than double the effect of age, and
their cumulative effect larger than the effect of market value?

H 0 : 1  2  2 and 1   2   4
 ECON 644 Class 6

Testing Hypotheses in Stata

. quietly regress salary ceoten age grad mktval

. test (ceoten age)

( 1) ceoten = 0
( 2) age = 0

F( 2, 172) = 1.99
Prob > F = 0.1397 Does Not Reject

. test (ceoten age grad mktval)

( 1) ceoten = 0
( 2) age = 0
( 3) grad = 0
( 4) mktval = 0

F( 4, 172) = 9.87
Prob > F = 0.0000 Rejects

. test (ceoten = 2*age) (ceoten+age=mktval)

( 1) ceoten - 2*age = 0
( 2) ceoten + age - mktval = 0

F( 2, 172) = 1.98
Prob > F = 0.1413 Does Not Reject
 ECON 644 Class 6

(2) Stata Scalars

. regress salary ceoten age grad mktval

Source | SS df MS Number of obs = 177

-------------+---------------------------------- F(4, 172) = 9.87
Model | 11348587.4 4 2837146.84 Prob > F = 0.0000
Residual | 49417377.4 172 287310.334 R-squared = 0.1868
-------------+---------------------------------- Adj R-squared = 0.1678
Total | 60765964.7 176 345261.163 Root MSE = 536.01

------------------------------------------------------------------------------
salary | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ceoten | 10.55413 6.019706 1.75 0.081 -1.327879 22.43614
age | 1.649817 5.162111 0.32 0.750 -8.539427 11.83906
grad | -43.43346 82.2966 -0.53 0.598 -205.8748 119.0079
mktval | .0371649 .0063706 5.83 0.000 .0245903 .0497395
_cons | 578.0711 288.3528 2.00 0.047 8.905383 1147.237
------------------------------------------------------------------------------

Note: The command regress comes with both r() and e()
scalars. They can be listed with the commands:
. return list
. ereturn list
 ECON 644 Class 6

Stata r() Scalars

. return list
scalars:
r(level) = 95
matrices:
r(table) : 9 x 5

Note: One scalar r(level), one matrix r(table). The scalar can be
listed again with the command:
. display r(level)
95

The matrix can be listed with the command:

. matrix list r(table)
r(table)[9,5]
ceoten age grad mktval _cons
b 10.554131 1.6498172 -43.433457 .03716492 578.07109
se 6.019706 5.1621111 82.296605 .00637059 288.35275
t 1.7532635 .31960125 -.52776731 5.8338252 2.0047358
pvalue .08133824 .74965849 .59834103 2.627e-08 .04655991
ll -1.3278789 -8.5394271 -205.87479 .02459032 8.9053834
ul 22.436141 11.839061 119.00787 .04973953 1147.2368
df 172 172 172 172 172
crit 1.9738522 1.9738522 1.9738522 1.9738522 1.9738522
eform 0 0 0 0 0
 ECON 644 Class 6

Stata e() Scalars

. ereturn list
scalars:
e(N) = 177
e(df_m) = 4
e(df_r) = 172
e(F) = 9.874851357165884
e(r2) = .1867589431214087
e(rmse) = 536.013370798667
e(mss) = 11348587.35367142
e(rss) = 49417377.39209128
e(r2_a) = .167846360403302
e(ll) = -1360.912292053917
e(ll_0) = -1379.207694429356
e(rank) = 5
macros:
e(cmdline) : "regress salary ceoten age grad mktval"
e(title) : "Linear regression"
e(marginsok) : "XB default"
e(vce) : "ols"
e(depvar) : "salary"
e(cmd) : "regress"
e(properties) : "b V"
e(predict) : "regres_p"
e(model) : "ols"
e(estat_cmd) : "regress_estat"
matrices:
e(b) : 1 x 5
e(V) : 5 x 5
functions:
e(sample)
 ECON 644 Class 6

Stata e() Scalars

Note: Twelve scalars e(), ten macros e(), two matrices e(), and
one function e().

A scalar can be listed again with the command:

. display e(F)
9.8748514

A matrix can be listed with the command:

. matrix list e(b)

e(b)[1,5]
ceoten age grad mktval _cons
y1 10.554131 1.6498172 -43.433457 .03716492 578.07109
 ECON 644 Class 6

(3) Unpacking the t-stat and F-stat Calculations

For the t-stats:

. display _b[ceoten]/_se[ceoten]
1.7532635

. display _b[age]/_se[age]
.31960125

. display _b[grad]/_se[grad]
-.52776731

. display _b[mktval]/_se[mktval]
5.8338252

The critical t-value:

. display invt(e(df_r),.975)
1.9738522
 ECON 644 Class 6

Unpacking the t-stat and F-stat Calculations

For the F-stat:
. display e(r2)
.18675894

. display e(df_m)
4

. display e(df_r)
172

Rur 2 / q
. display (e(r2)/e(df_m))/((1-e(r2))/e(df_r)) F
1  Rur 2  / (n  k  1)
9.8748514

The critical F-value

. display invF(e(df_m),e(df_r),.95)
2.4241928

Note: F-stat 9.87 is much larger than the critical F-value 2.42.
So, the test rejects that the model is insignificant (that all slope
coefficients are jointly zero).