用STATA做实证研究.pptVIP

* * * * * * * * * * * * * * * * * * * * * * * * * * Ramsey test: test functional form -ivreset, polynomial(#) small- Works only after -ivreg2- or -reg- * -sureg- : seemingly unrelated equations -sureg (piatMathStd1 hhAcc I_WAcc ) (hhAcc edu race gender) (I_WAcc edu race gender) * -nlsur-: -nlsur (piatMathStd1={xb_h}* hhDAcc +{xb_w}*I_WAcc +({xb_edu}* edu+{xb_race}*race+{xb_g}*gender +{xb_cons}*cons)*(1+{xb_xh}*hhDAcc+{xb_xw}*I_WAcc)) if quarter==20 * Post-estimation -est save filename, replace-; -gen sample=e(sample); -esttab, se nogap noline star(* 0.10 ** 0.05 *** 0.01) b(%8.3f)-; est2vec table_nlsur, replace vars(`indep) est2tex table_nlsur, path(g:\documents\teaching\stataworkshop\) preserve mark(starb) label levels(90 95 99) digits(3) replace -predict- and -predictnl- -test- and -testnl- -margins- * -probit-, -dprobit-, and -mprobit- -probit- estimates a probit model of a binary dependent variable. -dprobit- reports “marginal effects” (change in probability due to change in x) Example: Treat work and not work as a binary decision in each quarter -gen hD=0 -replace hD=1 if h0 -probit hD edu race gender quarter, vce(robust)- -dprobit hD edu race gender quarter, vce(robust)- * -mprobit- estimates a multinomial probit model where dependent variable is different alternatives Example: treat hD and I_W as four alternatives -egen c=group(hD I_W) -tab hd I_W, sum(c) * -mprobit- -mprobit c edu gender race quarter, vce(robust) Multinomial probit performs several integrals. Can be extremely slow.. * Panel Data Estimators: -xt- Note this is a long data of children’s age in quarters; -bysort childId: keep if childId==_N -xtreg-: when there is only one “cluster” -xtreg piatMathStd1 I_WSum hhDSum, fe i(momId); * Clustered model -xtmixed-: when there are more than one cluster for example, households living in different counties within states xtmixed piatMathStd1 hhDAcc I_WAcc || momId: mEdu || county: || state:, difficult; note -momId: mEdu- means mEdu affects decisions