Re: st: svy bootstrap and nlsur function evaluator program

[Stefano Verde <verdes@tcd.ie>]

Dear Stas,

I've microdata from a household budget survey, with the variable GF
which associates to each observation (i.e. household) a gross factor
(i.e. the number of households that observation represents in the
population).

I want to bootstrap some elastisticities, which are functions of the
estimated parameters from my demand system.

While the ordinary bootstrap works fine, I was hoping to do also a
weighted bootstrap to take account of the GF's.

I've been trying a number of combinations of -svyset- and -svy
bootstrap-, but no success so far.

In particular, when I type (note that "estimate" is just the name of
my program estimating the model)

svyset [iw=GF], bsr(GF)
svy bootstrap inc_1 = r(inc_1) inc_2 = r(inc_2) inc_3 = r(inc_3) inc_4
= r(inc_4) inc_5 = r(inc_5) inc_6 = r(inc_6) inc_7 = r(inc_7) ///
eta_1_1 = r(eta_1_1) eta_2_2 = r(eta_2_2) eta_3_3 = r(eta_3_3) eta_4_4
= r(eta_4_4)	eta_5_5 = r(eta_5_5) eta_6_6 = r(eta_6_6) eta_7_7 =
r(eta_7_7)	///
, reps(3): estimate if quint==1

I get this error

option reps() not allowed
r(198);

Why can't I spcecify the number of replications? So, the number of
replication is always the same with -svy bootstrap-? What is this
number? I couldn't find this information on the manuals.

If then the only change I make is to omit "reps(3)" all the rest
staying the same, I get this error

insufficient observations to perform bootstrap
r(459);

Now, this is odd to me because as I said before the bootstrap
perfectly works when I don't use svy and weights.

May I ask your advice?

Thanks for your previous repsonse,
Best wishes,
Stefano
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/