加载中…Lesson-5
(2021-03-02 19:23:47)*construct virtual data with
cycling grammar and conduct nbreg/poisson model
regression.
clear
forv i=1/10{
forv j=2010/2020{
set obs 1
generate stkid=`i'
generate year=`j'
generate
indcd=`i'+100
generate
y=round(10*runiform())
generate
x=runiform()
generate
x1=runiform()
generate
x2=runiform()
generate
x3=runiform()
generate
x4=runiform()
generate
x5=runiform()
save
data`i'`j'.dta,replace
clear
}
}
use
data12010.dta,clear
forv j=2011/2020{
append using
data1`j'.dta
}
forv i=2/10{
forv j=2010/2020{
append using
data`i'`j'.dta
}
}
save
datatest.dta,replace
forv i=1/10{
forv j=2010/2020{
erase
data`i'`j'.dta
}
}
use
datatest.dta,clear
br
replace y=1 if
y==0
global Model
"nbreg"
global DV "y"
global IV "x"
global Moderator
"x1"
global Control
"x3-x5"
center $IV
$Moderator
$Model $DV $IV
$Control
est store a1
$Model $DV $Moderator
$Control
est store a2
$Model $DV $IV $Moderator
$Control
est store a3
$Model $DV
c.c_$IV##c.c_$Moderator $Control
est store a4
outreg2 [a*] using
regression,excel dec(3) replace alpha(0.01,0.05,0.1)
logout,save(summary) excel
dec(3) replace:summarize $DV $IV $Moderator $Control
logout,save(cor) excel dec(3)
replace:pwcorr_a $DV $IV $Moderator $Control
*we can use another form of
dependent variable which is logarithm of
dv
*stating from the end of the
first sentence, under the logarithm precondition,we can use OLS
model
gen lny=ln(y)
reg lny c.c_$IV##c.c_$Moderator
$Control
*if we suspect independent
variable is endogenous we can solve it in two ways at
least
*one is use one year lagged
variables to fit endogenous variable to form an exogenous
variable
*another is to use averaged
industry variable excluding itself.
*solve endogeneity issue with
one-year lagged variable
tsset stkid year
gen lx1=l.x1
gen lx2=l.x2
gen lx3=l.x3
gen lx4=l.x4
gen lx5=l.x5
reg x lx1-lx5
predict xr1
center xr1
$Model $DV
c.c_xr1##c.c_$Moderator $Control
*solve endogenous issue with
average industry variable except itself.
*first of all i have to create
some industries same
replace indcd=101 if
indcd==110
replace indcd=102 if
indcd==109
replace indcd=103 if
indcd==108
replace indcd=104 if
indcd==107
replace indcd=105 if
indcd==106
gen n=1
bysort indcd year:egen
sumx=sum(x)
bysort indcd year:egen
sumn=sum(n)
gen
aveindx=(sumx-x)/(sumn-1)
reg x aveindx
x1-x5
predict xr2
center xr2
$Model $DV
c.c_xr2##c.c_$Moderator $Control
*we can also use poisson to
replace nbreg though it differs a little from nbreg
poisson $DV
c.c_$IV##c.c_$Moderator $Control
*if data is panel form we can
use xtnbreg or xtpoisson to conduct regression
xtset stkid year
eststo:xtnbreg $DV
c.c_$IV##c.c_$Moderator $Control
eststo:xtpoisson lny
c.c_$IV##c.c_$Moderator $Control
esttab,t(%4.3f)
b(%4.2f)
eststo clear
喜欢
0
赠金笔
前一篇：清平乐·万物悄然重生
后一篇：风动夜泉城