Monte-Carlo simulation error:"no obeservations"

[Pieter Janssens]

Hello,

I am new to Stata and for a project I have to peform a monte carlo simulation with 500 replications using three different sample sizes, being $N_1$ = 50, $N_2$ = 200 and $N_3$ = 2000

Given: y1 = b0 + b1*y2 + b2*x1 + b2*x2
and y2 = a0 + a1*x1 + a2*x1+a2*x2 + a3*z1 + a4*z2+a5*z3+a6*z6


where y2 is the endogenious variable.

Below is what I have programmed so far. The problem is that I keep getting the error "no observations". Though, I really do not see where the problem is as I have almost no experience in Stata. Can somebody please help me? Many thanks in advance!


tempfile mc30 mc100 mc200 mc1600

foreach n in 50 200 2000 {
local j = 1
while `j' <= 500 {
quietly {
preserve
sample `n', count
noisily di in yellow "." _continue


gen x1= 2*uniform()
gen x2 = rnormal(0,1)
gen z1 = rnormal(4,10)
gen z2 = (invnorm(uniform()))^2
gen z3 = rnormal(2,3)
gen z4 = rnormal(2,3)
gen e = rnormal(0,1)
gen v = rnormal(0,1)

gen y2_50 = 1 + 1*x1 +1*x2+1*z1+1*z2+e
gen y1_50 = 1 + y2_50 + x1 +x2 + v
reg y2_50 x1 x2 z1 z2
reg y1_50 x1 x2 y2_50
matrix b1=e(b)
svmat b1, name(ols) // save the regression coefficients and collapse in
// order to only have one observation per coefficient

collapse (mean) ols*

gen N = `n'

if `j' > 1 {
append using `mc`n''
}

save `mc`n'', replace
restore
local j = `j' + 1
}
}
di "`n'"
}

use `mc50', clear
append using `mc100'
append using `mc200'
append using `mc2000'
// here we append the four mc simulations (each with 1000 observations for B1
// and B2) of four different sample sizes together to compare them in task 2.

[Stephen Jenkins]

Have you looked at

help simulate

[Pieter Janssens]

Dear, Mr. Jenkins,

Yes I have looked at it, but it did not give me a solution to my question.

[Stephen Jenkins]

I can't comment further with the information provided so far. I am confident that simulate would be a route to undertaking MC analysis, and don't understand why it is not part of your routine. Moreover, you have not provided sufficient information about your (simulated) data sets at each step in what you do -- so how can anyone tell where the problem might arise? Remember, as the Forum FAQ advises, (i) show us exactly what you typed and exactly what Stata gave back, and (ii) report all this using CODE delimiters (the use of which is easy and also explained in the FAQ. So one way to possibly get advice would be provide further information, e.g. including descriptive statistics and other output that is relevant to checking sample sizes