Monte Carlo Simulation and Kernel density graphing

[Amy Murphy]

Hi

I'm stuck with a monte carlo simulation for testing 5 different models and comparing them at the end.

This is my current code: I would be extremely grateful for any help as unsure why it isn't working

** Generate the observable treatment effect
gen `Ty0' = gamma_0 + gamma_1* Tage+ gamma_2*Tfemale+ gamma3*Tage*Tfemale+Terror

gen yobs= Ty0 + TW

reg Ty0 TW
** Store value of beta
mat beta 'j' [`i',1] = ss1[1,1]

reg Ty0 TW Tage Tfemale Tfemale*Tage
** Store value of beta
mat beta 'j' [`i',2] = ss2[1,1]

reg Ty0 TW i.Tage
** Store value of beta
mat beta 'j' [`i',3] = ss3[1,1]

logit `TW' `Tage'
predict `pred_prob1' //
gen `lad'=1/`pred_prob1'^`TW'*(1-`pred_prob1'^(1-`TW'))
reg Ty0 Tage `lad' [W=`lad']
mat beta 'j' [`i',4] = ss4[1,1]

logit `TW' i.`Tage'
predict `pred_prob2' //
gen `lad2'=1/`pred_prob2'^`varw'*(1-`pred_prob2'^(1-`varw'))
reg `y0' `vage' `lad2' [w=`lad2']
mat beta 'j' [`i',5] = ss5[1,1]
}
end

** Run the programme R=1000 times
mat def beta1 = J(1000,2,.)
set seed 12345
forvalues i = 1/1000 {
SIM 'i' 5
}

svmat beta1

** Summarize each estimators
sum beta11 beta12 beta13 beta14 beta15

[Maarten Buis]

What does "isn't working" mean in your case?

Maybe the simulate command helps?

[Nick Cox]

In the first line of code

gen `Ty0' = gamma_0 + gamma_1* Tage+ gamma_2*Tfemale+ gamma3*Tage*Tfemale+Terror

you refer to a local macro Ty0. That will work if the local macro has been defined earlier in the code and contains a legal new variable name, but not otherwise.

Then you refer to Ty0 as a variable name. but commands using it depend on it having been created earlier.

This two puzzles are perhaps one and the same puzzle, so that the first command should refer to Ty0.

[Amy Murphy]

Thankyou both Maarten and Nick!

I tried the SIM code and got a loop together, specifying y0 like you suggested! Worked for my easier regressions so thankyou muchly! However now i see the 'unknown weight type' error in my logit model. I have coded :
logit `TW' `Tage'
predict `pred_prob1'
gen `lad'= 1/((`pred_prob1'^`TW')*((1-`pred_prob1')^(1-`TW')))
reg `Tyobs' `Tage' [W=`lad']
[pweight=W]
*mat beta`j'[`i',4] = _b[`TW']

but not sure where it has gone worng