Week6 Do File

* 6 Issues in linear regressions

set mem 100m

set matsize 4000

*log using e:\0Teaching\1Empirical\log120413.smcl

use e:\0Teaching\1Empirical\Data_Compustat2006_2008.dta, clear

gen cusip8=substr(cusip,1,8)

label var cusip8 "8-digit CUSIP"

ta fyear

rename fyear year

drop if year==2005

// merge with market cap data

merge 1:1 cusip8 year using "e:\0Teaching\1Empirical\mktcap2006_2008.dta"

keep if _merge==3

drop _merge

* Generating Tobin's q

mvencode pstkl, mv(0) override

egen mve=rowmean(q1mc-q4mc)

gen qcp=(mve+pstkl+lct-act)/at

sum qcp, det

gen qgim=(at+mve-seq-txdb)/at

sum qgim, det

gen sic1=substr(sic,1,1)

destring gvkey, replace

merge 1:1 gvkey year using "e:\0Teaching\1Empirical\entropy2006_2008.dta"

keep if _merge==3

drop _merge

ta nseg

sum entropy, det

pwcorr qcp qgim entropy, st(10) sig

// Generating log variables

gen lat=log(at)

hist lat, norm

gen lsale=log(1+sale)

sum emp, det

gen lemp=log(1+emp)

mvencode xad, mv(0) override

mvencode xrd, mv(0) override

hist xad

gen lxad=log(1+xad)

hist lxad, norm

hist xrd

gen lxrd=log(1+xrd)

hist lxrd, norm

// Generating ratios

gen roa=ni/at

sum roa, det

hist roa

drop if roa<-1 | roa>1

hist roa

gen cr=act/lct

sum cr, det

gen dr=lt/at

sum dr, det

gen adint=xad/sale

sum adint, det

gen rndint=xrd/sale

sum rndint, det

drop if rndint>4

* xtset

*xtset gvkey year

egen firmid=group(gvkey)

xtset firmid year

* Set the omitted category

char _dta[omit] "prevalent"

* OLS - baseline

xi: reg roa lat cr dr adint rndint entropy i.sic1 i.year

predict yhat, xb

est store baseline

outreg2 using "e:\0Teaching\1Empirical\week6_base.out", bdec(4) tstat excel 2aster bracket adjr2 replace

*Testing the size of coefficients

test adint==rndint

* Check for multicollinearity & heteroskedasticity

estat vif

estat hettest

* Standardized coefficients

xi: reg roa lat cr dr adint rndint entropy i.sic1 i.year, beta

outreg2 using...