sampling in group not working properly

[xinyongtian]

summary, the example dataset contain 2 columns, personID and CaseID. one perosnID has multi CaseID.

The purpose is to sample one case for each person. the result is not correct

but if replace with any of two codes marked with #$$$$$$ ,i.e.1.each person has same number of cases

or 2. sample .N instead of directly sampling CaseID, the result is correct.

library(data.table)
#1. create dataset

set.seed(1222)

weight=sample(c(0.1,0.2,0.7),100,replace=T)
dt=data.table(personID=sample(1:100,500,replace=T,prob=weight), CaseID=1:500)

dt=data.table(personID=rep(1:100,5), CaseID=1:500)#$$$$$$$$$$ no weight works

table(dt$personID)
dt
#2.random sampling one case per person

set.seed(1222)

s=dt[,list(CaseID=sample(CaseID,1)),by=personID];s

s=dt[,list(CaseID=CaseID[sample(.N,1)]),by=personID];s #$$$$$$$$$$ sample N works

#3. check

library(sqldf)
sqldf("select * from s except select * from dt") #rows that not in origninal dataset.