Subject: Biost 517: Homework #6 bootstrapping in Stata
The Stata commands given in class for bootstrapping the median need to be
modified: centile is not returning the median appropriately in the _result()
variable.
In order to get the standard error of the sample median for spd12 in the dose 0
group (for example):
1. First get a data set that only contains the dose 0 group (this is necessary
because the bootstrap resamples from all cases)
drop if dose > 0
drop if spd12==.
2. Now bootstrap to get standard error of the sample median. I am now going to
recommend that you use summarize, detail, in which case the 10th result
corresponds to the sample median. So the Stata command is:
bs "summ spd12, detail" "_result(10)", reps(1000)
(Note that every time you execute this, you will get a slightly different
answer, because every bootstrap is a different batch of pseudosamples. Hence,
do not be surprised if your answer does not agree exactly with other peoples.
Just use the standard errors you get.)
You will have to repeat this process with the other dose group and with the
change in spd over the treatment period.
Scott