我想用预处理RMSPE省略安慰剂，这比我的治疗单位高2倍。 但是，当我使用带有pre_limit_mult（2）的synth_runner命令并完全按照他们在论文中给出的示例时，每个安慰剂仍然包含在图表中。有谁知道问题可能是什么？
答案 0 :(得分：1)
以下是MCVE的示例。
如帮助文件中所述，如果该状态的匹配质量（通过治疗前根均方预测误差测量），则pre_limit_mult(K)将从推理中排除安慰剂状态，大于K倍处理单元的匹配质量。这不会改变哪些安慰剂线被绘制，它只会降低p值。
但是，你可以放弃那些你不想在图中显示的那样：
use "smoking.dta", clear
tsset state year
local K = 2 // pre_limit_mult factor
/* all 38 placebos */
synth_runner cigsale beer(1984(1)1988) lnincome(1972(1)1988) retprice age15to24 cigsale(1988) cigsale(1980) cigsale(1975), trunit(3) trperiod(1989) gen_vars
di e(n_pl) // The number of placebo averages used for comparison
di e(avg_pre_rmspe_p) // The proportion of placebos that have a pre-treatment RMSPE at least as large as the average of the treated units
single_treatment_graphs
graph export "raw_all.pdf", name(raw) replace
drop pre_rmspe post_rmspe lead effect cigsale_synth
/* same as above, but p-values change when we exclude 17 states */
synth_runner cigsale beer(1984(1)1988) lnincome(1972(1)1988) retprice age15to24 cigsale(1988) cigsale(1980) cigsale(1975), trunit(3) trperiod(1989) gen_vars pre_limit_mult(`K')
di e(n_pl) // The number of placebo averages used for comparison
di e(avg_pre_rmspe_p) // The proportion of placebos that have a pre-treatment RMSPE at least as large as the average of the treated units
single_treatment_graphs
graph export "raw_all2.pdf", name(raw) replace
/* exclude 17 states from graph */
preserve
sum pre_rmspe if state=="California":state, meanonly
gen in_sample = cond(pre_rmspe < = r(mean)*`K',1,0)
distinct state if in_sample == 1 // CA is counted as in sample, so this matches 21 above
drop if in_sample == 0
single_treatment_graphs
graph export "raw_limited.pdf", name(raw) replace
restore
另一种方法是为图形命令编写一个包装器，为您执行此操作并接受if选项。如果你计划做很多事情，可能值得你努力。
您可能需要ssc install distinct，因为这是用户编写的命令。