[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

[no subject]

g	mean of 100 estimates
0.4	-3.18674  
0.3	-3.083586  
0.2	-3.056365  
0.1	-2.994547  
0.05	-2.973532

The divergence may not seem large, but it is consistent for a number of
simulations, and values of nip, so it appears to be a problem.   

My question is whether its due to my data set up, my implementation of
-gllamm- command, or something else.



(PS  My apologies to those who can write elegant/efficient code)

* in the model below there is a 'true' observation of the latent
variable x
* z1-z4 are the 4 measurements of that variable for each individual, 
* y is the observed response
* -gllamm- attempts to retrieve the known parameters given only the
* of y and z1-z4
* values reported above are the mean of 100 replicates for the 5 values
of g

set matsize 500
matrix result=J(500,6,0)

set more off

foreach num of numlist 1/5 {
foreach iter of numlist 1/100 {

drop *
set obs 500

gen g=0.05
replace g=0.1 if `num'==2
replace g=0.2 if `num'==3
replace g=0.3 if `num'==4
replace g=0.4 if `num'==5

gen err=invnorm(uniform())
gen x=uniform()-0.5
gen ys=1-3*x+err
gen y=0
replace y=1 if ys>0

gen z1=x+g*invnorm(uniform())
gen z2=x+g*invnorm(uniform())
gen z3=x+g*invnorm(uniform())
gen z4=x+g*invnorm(uniform())

gen id=_n
expand 5
sort id
by id: gen t=_n

gen resp=z1
replace resp=z2 if t==2
replace resp=z3 if t==3
replace resp=z4 if t==4
replace resp=y if t==5

gen op1=0
replace op1=1 if t==1
gen op2=0
replace op2=1 if t==2
gen op3=0
replace op3=1 if t==3
gen op4=0
replace op4=1 if t==4

gen yvar=0
replace yvar=1 if t==5

gen var=1
replace var=2 if t==5

eq i1: op1 op2 op3 op4 yvar

gllamm resp  yvar op1 op2 op3 op4 , nip(20) nocons i(id) nrf(1) eqs(i1)
link (ident probit) family(gauss binom) lv(var) fv(var) trace adapt

matrix t=e(b)

lincom [id1_1l]_b[yvar]

matrix result[(`num'-1)*100+`iter',1]=t[1,10]
matrix result[(`num'-1)*100+`iter',2]=r(se)
matrix result[(`num'-1)*100+`iter',3]=(exp(t[1,6]))^2
matrix result[(`num'-1)*100+`iter',4]=`num'

matrix list result

Dr Michael Burton
Associate Professor 
School of Agricultural & Resource Economics (MO89)
Faculty of Natural & Agricultural Sciences
University of Western Australia
35, Stirling Highway
Crawley (Perth), WA 6009 Australia
tel +61 8 6488 2531
fax +61 8 6488 1098
The University of Western Australia:  CRICOS Provider No. 00126G

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index