- /*Next choose represantitive treatment-level*/
- matrix tp=J(101,1,0)
- scalar j=0
- forvalues i=1/101{
- mat tp[`i',1]=j
- scalar j=j+0.01
- }
- /*Cutpoints*/
- gen cut=0.031 if treat<=0.031
- replace cut=0.068 if treat>0.031 & treat<=0.068
- replace cut=0.124 if treat>0.068 & treat<=0.124
- replace cut=0.440 if treat>0.124 & treat<=0.440
- replace cut=1 if treat>0.440 & treat<=1
- doseresponse2 lgross_value lnk tfp finan age ex RD subsidy NZ _I*, outcome(export_avg) ///
- t(treat) gpscore(gps) predict(t_hat) sigma(sd) cutpoints(cut) index(mean) ///
- nq_gps(4) dose_response(dose_res) reg_type_t(cubic) reg_type_gps(cubic) ///
- family(bin) link(logit) tpoints(tp) bootstrap(yes) boot_reps(100) analysis(yes) ///
- analysis_level(0.95) filename("output_avg_cubic_boostap_2010") ///
- graph("graph_avg_cubic_boostrap_2010") detail
以下是我跑出的模型结果Fractional Logit1回归结果
Fractional Logit2回归结果
最后再附一下那篇英文文献的平衡性检验,其使用的是LotteryDataSet.dta这个文件