problem with simulate command and rclass

[Andrew Xavier]

Dear all,

I'm having a problem with simulate command. I have written an r-class program. When I run the program and use "return list", I get the estimates I need. However, when I use simulate to store those estimates, simulate returns an error and it does not store those values. I have tried using other estimates and it works fine but when trying to extract pval and ub it simply does not work. To try to solve the problem, I tried to extract the result, store it in a local, then generate a variable with that value and then use the r(mean) value of that variable - however this does not work. This is really odd and spent hours trying to figure out why it does not work without success.

You can find below the code. Any help will be very much appreciated.

ssc install survsim
cap prog drop sim_model
prog define sim_model, rclass
    clear
    set obs 300
    gen trt = rbinomial(1,0.5)
    survsim time fail, lambdas(0.07) distribution(exponential) covariates(trt -.28768207) maxtime(48) 
    stset time, fail(fail)
    gen cured = rbinomial(1,0.20)
    replace time=48 if cured==1
    replace fail=0 if cured==1
    drop cured
    stset time, fail(fail)

    fmm: (pointmass fail) (streg trt, distribution(exponential))
    mat a=r(table_m2)
    local pval=el(a,4,1)
    gen pval=`pval'
    sum pval
    return scalar pval=r(mean)
    local pval=r(mean)
        
    local ubln=el(a,6,1)
    local ub=exp(`ubln')
    gen ub=`ub'
    sum ub
    return scalar ub=r(mean)
    local ub=r(mean)    
    ********************
    gen reject_ub=.
    replace reject_ub=1 if `ub'<1
    replace reject_ub=0 if reject_ub==.
    sum reject_ub    
    return scalar reject_ub=r(mean)

    gen reject_pvalue=.
    replace reject_pvalue=1 if `pval'<0.05
    replace reject_pvalue=0 if reject_pvalue==.
    sum reject_pvalue
    return scalar reject_pvalue=r(mean)

end

sim_model
return list

simulate ub=r(ub) pval=r(pval) reject_pvalue=r(reject_pvalue) reject_ub=r(reject_ub), reps(5): sim_model

[Joro Kolev]

I do not see anything wrong on first look.

Try -set trace on- and report exactly what error message -simulate- returns.

You can also try whether it works for only one returned result, e.g., just ub.

[Fei Wang]

I don't know why, but the code works well only when I turn on the "trace" option as below.

simulate ub=r(ub) pval=r(pval) reject_pvalue=r(reject_pvalue) reject_ub=r(reject_ub), trace reps(5): sim_model

[Andrew Xavier]

I do not see anything wrong on first look.

Try -set trace on- and report exactly what error message -simulate- returns.

You can also try whether it works for only one returned result, e.g., just ub.

I used set trace on and there is nothing wrong. Simulate does not return an error

[Andrew Xavier]

I don't know why, but the code works well only when I turn on the "trace" option as below.

simulate ub=r(ub) pval=r(pval) reject_pvalue=r(reject_pvalue) reject_ub=r(reject_ub), trace reps(5): sim_model

Thank you Fei Wang. This solved the problem! I wonder what could be the problem here?

[Fei Wang]

Thank you Fei Wang. This solved the problem! I wonder what could be the problem here?

Sorry Andrew, I don't know why. Turning on and off "trace" is not supposed to affect the essence of the code. Hope some experts can decipher.

[Andrew Xavier]

Sorry Andrew, I don't know why. Turning on and off "trace" is not supposed to affect the essence of the code. Hope some experts can decipher.

Let's see if someone can decipher it

[Joro Kolev]

This does not sound regular. Ask Stata technical support, and report what they said.

[Andrew Xavier]

Thanks Joro Kolev,

I emailed them about this query and they will look at my query next week. I will then report what they said here

[Andrew Xavier]

I received a reply from STATA:

You need to get the values from -r(table)-, which is the documented result. Results from the non-documented -r(table_m2)- are not posted when the command is called quietly.

i have done this and the simulate command works fine

Thanks