[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Problems with psmatch
I have a problem when trying to replicate a psmatch estimation.
I have been setting the seed instruction to the same number to assure
that the results after calculating the confidence intervals in psmatch
(with kernel and bstrap options) were replicable. However I have found
out that the results may differ even if I use the same seed under
1. If I execute the same psmatch instruction twice in the same stata
session and before each run we set the same seed, the results differ.
2. If we do the following experiment: open two stata sessions at the
same time. In the first session we run a bstrap instruction after
setting the seed. In the second session before runing exactly the same
sequence of instructions we decide to run some seemingly unconnected
instructions (like graphing a kdensity or running a hotelling test) the
results again differ.
Do you have a clue about what may be happening?
How can I be sure that all internal variables have been cleared before
running a certain instruction?
* For searches and help try: