<span class='text_page_counter'>(1)</span><div class='page_container' data-page=1>
FUNCTIONAL FORMS
Truong Dang Thuy
</div>
<span class='text_page_counter'>(2)</span><div class='page_container' data-page=2>
Linear model
Consider a linear regression function
: change in Y when X increases by 1 unit.
Sometimes the relationship is not linear.
Common functional form:
Log-linear
Log-lin
Lin-log
Reciprocal
Polynomial
0
1
<i>Y</i>
<i>X</i>
</div>
<span class='text_page_counter'>(3)</span><div class='page_container' data-page=3>
Functional forms
<b>Linear model </b>
<b>Log-linear </b>
<b>Lin-log </b>
<b>Log-lin </b>
0 1
<i>Y</i>
<i>X</i>
0 1
ln
<i>Y</i>
ln
<i>X</i>
0 1
ln
<i>Y</i>
<i>X</i>
0 1
</div>
<span class='text_page_counter'>(4)</span><div class='page_container' data-page=4>
Functional forms
<b>Reciprocal (negative beta) </b>
<b>Reciprocal (positive beta) </b>
0 1 1
1
0
<i>Y</i>
<i>X</i>
1
0 1 1
1
0
<i>Y</i>
<i>X</i>
</div>
<span class='text_page_counter'>(5)</span><div class='page_container' data-page=5>
Example dataset
Viet Nam Provincial data on (file ‘
gdpprov.xlsx
’)
gdp
:
provincial GDP (mil. VND)
labfo
:
number of laborers of provinces (1000
persons)
</div>
<span class='text_page_counter'>(6)</span><div class='page_container' data-page=6>
<b>Record of </b>
<b>commands </b>
<b>Record of results </b>
<b>Variables </b>
<b>(data) </b>
<b>Commands </b>
<b>Taskbar </b>
</div>
<span class='text_page_counter'>(7)</span><div class='page_container' data-page=7>
Import data
<b>Copy from Excel </b>
</div>
<span class='text_page_counter'>(8)</span><div class='page_container' data-page=8>
Data description
</div>
<span class='text_page_counter'>(9)</span><div class='page_container' data-page=9>
Linear function
</div>
<span class='text_page_counter'>(10)</span><div class='page_container' data-page=10>
LOG-LINEAR MODEL
The Cobb-Douglas Production Function:
can be transformed into a linear model by taking natural
logs of both sides:
The slope coefficients can be interpreted as elasticities.
<i><sub>If (B</sub></i>
<i><sub>2</sub></i>
<i><sub> + B</sub></i>
<i><sub>3</sub></i>
<sub>) = 1, we have constant returns to scale. </sub>
<i><sub>If (B</sub></i>
<i><sub>2</sub></i>
<i><sub> + B</sub></i>
<i><sub>3</sub></i>
<sub>) > 1, we have increasing returns to scale. </sub>
<i><sub>If (B</sub></i>
<i><sub>2</sub></i>
<i><sub> + B</sub></i>
<i><sub>3</sub></i>
<sub>) < 1, we have decreasing returns to scale. </sub>
3
2
1
<i>B</i>
<i>B</i>
<i>i</i>
<i>i</i>
<i>i</i>
<i>Q</i>
<i>B L K</i>
1
2
3
</div>
<span class='text_page_counter'>(11)</span><div class='page_container' data-page=11>
Log-linear model
_cons 3.06333 .4515804 6.78 0.000 2.174233 3.952426
linvest .644785 .0405325 15.91 0.000 .5649824 .7245876
llabor .508612 .0643267 7.91 0.000 .381962 .635262
lgdp Coef. Std. Err. t P>|t| [95% Conf. Interval]
Total 224.910559 270 .833002069 Root MSE = .42886
Adj R-squared = 0.7792
Residual 49.2915017 268 .183923514 R-squared = 0.7808
Model 175.619057 2 87.8095284 Prob > F = 0.0000
F( 2, 268) = 477.42
Source SS df MS Number of obs = 271
. reg lgdp llabor linvest
(17 missing values generated)
. gen linvest = ln(rinvest)
. gen llabor = ln(labfo)
(10 missing values generated)
. gen lgdp = ln(rgdp)
</div>
<span class='text_page_counter'>(12)</span><div class='page_container' data-page=12>
LOG-LIN OR GROWTH MODELS
The rate of growth of real GDP:
can be transformed into a linear model by taking natural logs
of both sides:
<i>Letting B</i>
<sub>1</sub>
<i> = ln RGDP</i>
<sub>0</sub>
<i> and B</i>
<sub>2</sub>
<i> = ln (l+r), this can be </i>
rewritten as:
<i>ln RGDP</i>
<sub>t</sub>
<i> = B</i>
<sub>1</sub>
<i> +B</i>
<sub>2</sub>
<i> t </i>
<i>B</i>
<i><sub>2</sub></i>
<i> is considered a semi-elasticity or an instantaneous growth rate. </i>
<i>The compound growth rate (r) is equal to (e</i>
<i>B2</i>
<i> – 1). </i>
0
(1
)
<i>t</i>
<i>t</i>
<i>RGDP</i>
<i>RGDP</i>
<i>r</i>
0
</div>
<span class='text_page_counter'>(13)</span><div class='page_container' data-page=13>
LOG-LIN MODEL
t 290 3 1.416658 1 5
Variable Obs Mean Std. Dev. Min Max
. sum t
</div>
<span class='text_page_counter'>(14)</span><div class='page_container' data-page=14>
LOG-LIN MODEL
</div>
<span class='text_page_counter'>(15)</span><div class='page_container' data-page=15>
LIN-LOG MODELS
Lin-log models follow this general form:
<i>Note that B</i>
<i><sub>2</sub></i>
<i> is the absolute change in Y responding to a </i>
<i>percentage (or relative) change in X </i>
<i>If X increases by 100%, predicted Y increases by B</i>
<i><sub>2</sub></i>
units
1
2
ln
<i>i</i>
<i>i</i>
<i>i</i>
</div>
<span class='text_page_counter'>(16)</span><div class='page_container' data-page=16>
Exercise – lin-log model
Data: from VHLSS 2010
income
: individual annual income (1000 VND)
healthcost
: individual annual cost for health care
(1000 VND)
Use the data in ‘healthcost.dta’ to run the
regression
where
hcshare
is the share of health cost in income.
0
1
ln
</div>
<span class='text_page_counter'>(17)</span><div class='page_container' data-page=17>
Health cost with Lin-log model
_cons .421608 .0322026 13.09 0.000 .35847 .484746
lincome -.0341629 .0029364 -11.63 0.000 -.0399202 -.0284056
hcshare Coef. Std. Err. t P>|t| [95% Conf. Interval]
Total 75.7996618 3474 .021819131 Root MSE = .14494
Adj R-squared = 0.0372
Residual 72.9563097 3473 .021006712 R-squared = 0.0375
Model 2.84335206 1 2.84335206 Prob > F = 0.0000
F( 1, 3473) = 135.35
Source SS df MS Number of obs = 3475
. reg hcshare lincome
. gen lincome = ln(income)
</div>
<span class='text_page_counter'>(18)</span><div class='page_container' data-page=18>
RECIPROCAL MODELS
Lin-log models follow this general form:
Note that:
<i>As X increases indefinitely, the term approaches zero and Y approaches </i>
<i>the limiting or asymptotic value B</i>
<i><sub>1</sub></i>
.
The slope is:
<i>Therefore, if B</i>
<sub>2</sub>
<i> is positive, the slope is negative throughout, and if B</i>
<sub>2</sub>
is negative,
the slope is positive throughout.
1
2
1
(
)
<i>i</i>
<i>i</i>
<i>i</i>
<i>Y</i>
<i>B</i>
<i>B</i>
<i>u</i>
<i>X</i>
2
1
(
)
<i>i</i>
<i>B</i>
<i>X</i>
2
<sub>2</sub>
1
(
)
<i>dY</i>
<i>B</i>
</div>
<span class='text_page_counter'>(19)</span><div class='page_container' data-page=19>
Exercise – Reciprocal model
Use the data in ‘
healthcost.dta
’ to run the
regression
0
1
1
<i>hcshare</i>
<i>income</i>
</div>
<span class='text_page_counter'>(20)</span><div class='page_container' data-page=20>
Exercise – Reciprocal model
_cons .023971 .0032251 7.43 0.000 .0176478 .0302943
invincome 942.4843 81.65964 11.54 0.000 782.3786 1102.59
hcshare Coef. Std. Err. t P>|t| [95% Conf. Interval]
Total 75.7996618 3474 .021819131 Root MSE = .14498
Adj R-squared = 0.0367
Residual 72.9997153 3473 .02101921 R-squared = 0.0369
Model 2.79994649 1 2.79994649 Prob > F = 0.0000
F( 1, 3473) = 133.21
Source SS df MS Number of obs = 3475
. reg hcshare invincome
</div>
<span class='text_page_counter'>(21)</span><div class='page_container' data-page=21>
POLYNOMIAL REGRESSION MODELS
The following regression predicting GDP is an example of a
quadratic function, or more generally, a second-degree
<i>polynomial in the variable time: </i>
The slope is nonlinear and equal to:
Exercise: run the above model with ‘gdpprov.dta’
2
1
2
3
<i>t</i>
<i>t</i>
<i>RGDP</i>
<i>A</i>
<i>A time</i>
<i>A time</i>
<i>u</i>
2
2
3
<i>dRGDP</i>
<i>A</i>
<i>A time</i>
</div>
<span class='text_page_counter'>(22)</span><div class='page_container' data-page=22>
SUMMARY OF FUNCTIONAL FORMS
<b>MODEL </b>
<b>FORM </b>
<b>SLOPE </b>
<b>ELASTICITY </b>
<b>(</b>
<i>dY</i>
<i>dX</i>
<b>) </b>
.
<i>dY X</i>
<i>dX Y</i>
Linear
<i>Y =B</i>
<i>1</i>
<i> + B</i>
<i>2</i>
<i> X </i>
<i>B </i>
2 2
(
)
<i>Y</i>
<i>X</i>
<i>B</i>
Log-linear
<i>lnY =B</i>
<i>1</i>
<i> + ln X </i>
2
(
)
<i>Y</i>
<i>B</i>
<i>X</i>
<i>B </i>
2
Log-lin
<i>lnY =B</i>
<i>1</i>
<i> + B</i>
<i>2</i>
<i> X </i>
<i>B Y </i>
2
( )
<i>B</i>
2
(
<i>X</i>
)
Lin-log
<i>Y</i>
<i>B</i>
1
<i>B</i>
2
ln
<i>X</i>
2
1
(
)
<i>B</i>
<i>X</i>
)
1
(
2
<i>Y</i>
<i>B</i>
Reciprocal
1 2
1
(
)
<i>Y</i>
<i>B</i>
<i>B</i>
<i>X</i>
<i>B</i>
<sub>2</sub>
(
1
<sub>2</sub>
)
<i>X</i>
<sub>2</sub>
(
1
)
<i>XY</i>
<i>B</i>
2
ln
</div>
<span class='text_page_counter'>(23)</span><div class='page_container' data-page=23>
COMPARING ON BASIS OF R
2
We cannot directly compare two models that have
different dependent variables.
We can transform the models as follows and compare RSS:
<i>Step 1: Compute the geometric mean (GM) of the dependent </i>
<i>variable, call it Y</i>
*
<sub>. </sub>
<i>Step 2: Divide Y</i>
<i><sub>i</sub></i>
<i> by Y</i>
*
to obtain:
<i>Step 3: Estimate the equation with lnY</i>
<i><sub>i</sub></i>
as the dependent variable
<i>using in lieu of Y</i>
<i><sub>i</sub></i>
as the dependent variable (i.e., use ln as the
dependent variable).
<i>Step 4: Estimate the equation with Y</i>
<i><sub>i</sub></i>
as the dependent variable
<i>using as the dependent variable instead of Y</i>
<i><sub>i</sub></i>
.
<i>i</i>
<i>i</i>
<i>Y</i>
<i>Y</i>
<i>Y</i>
~
*
<i>i</i>
<i>Y</i>
~
<i>Y</i>
~
<i><sub>i</sub></i>
<i>i</i>
</div>
<span class='text_page_counter'>(24)</span><div class='page_container' data-page=24>
MEASURES OF GOODNESS OF FIT
<i>R</i>
2
: Measures the proportion of the variation in the regressand
explained by the regressors.
<i>Adjusted R</i>
2
: Denoted as , it takes degrees of freedom into account:
Akaike’s Information Criterion (AIC): Adds harsher penalty for adding
more variables to the model, defined as:
<i>The model with the lowest AIC is usually chosen. </i>
Schwarz’s Information Criterion (SIC): Alternative to the AIC criterion,
expressed as:
<i><sub>The penalty factor here is harsher than that of AIC. </sub></i>
2
<i>R</i>
_
2 2
1
1 (1
)
<i>n</i>
<i>R</i>
<i>R</i>
<i>n k</i>
2
ln
<i>AIC</i>
<i>k</i>
ln(
<i>RSS</i>
)
<i>n</i>
<i>n</i>
ln
<i>SIC</i>
<i>k</i>
ln
<i>n</i>
ln(
<i>RSS</i>
)
<i>n</i>
<i>n</i>
</div>
<!--links-->