set seed 12345
gen random=runiform()
sort random
gen validation=_n<=450            //assume you sample size is 900
label define valid 1 "model_build" 0 "model_valid"
label val validation valid

感谢楼上！

我摸索了一个这样的
set obs 900
gen x=rnormal()
xtile x2=x, n(2)
然后导出数据到Mplus，加入一条if x2==1，或者if  x2==2

这样子可以吗？这样是随机分半了吗？

ps：怎么样将数据split成两个stata数据文件？

help bsample

Your method may be Ok but it's not a standard procedure, especially the use of "rnormal()".  Personally I have never seen Stata experts use "rnormal()" instead of "runiform()" for the aim of random sampling.

As regards to splitting the Stata file, there are several methods, the simplest one is to find and install the command "savesome"

Good luck

如果我在上面原有的语句里面，用runiform()替换rnormal(),这样做没问题吧？
谢谢！