'This code is provided 'as is'. There is no guarantee that it is free of errors. 

'Simulating critical values for the GLS Dickey-Fuller test of Elliott et al. (1996, Econometrica)

'In its current form, this code considers the trend model. However, it can be modified quite easily to consider the intercept model

'The number of replications
!reps = 100000

'Setting the sample size
!sample=150

'The sample used in the simulation:  an additional observation used to allow for a discarded initial value
!obs=!sample+1

'Creating a workfile to house the results
wfcreate GLS_!reps u !obs

'Creating vectors to store the test statistics (tests) and the critical values (cvs)
vector(!reps) tests
vector(3) cvs

'All observations
smpl 1 !obs

'Fixing the seed so results can be replicated
rndseed 1

'Creating c_bar and alpha_bar:  note that c_bar=-7 for the intercept model and c_bar=-13.5 for the trend model
scalar c_bar=-13.5
scalar alpha_bar=1+(c_bar/(!obs))

'The simulation loop begins
for !k=1 to !reps

'Creating the error term
series u=nrnd
genr trnd=@trend+1

'Initialisation
smpl 1 1 
series y = 0
genr ya=y
genr ca=1
genr ta=1

'Creating terms over the rest of the sample
smpl 2 !obs
series y = y(-1)+u
genr ya=y-alpha_bar*y(-1)
genr ca=1-alpha_bar
genr ta=trnd-alpha_bar*(trnd-1)

'Removing the effect of deterministics and creating the adjusted series and its first difference
smpl @all
equation eq1.ls ya ca ta
genr yd=y-(eq1.@coefs(1))-(eq1.@coefs(2)*trnd)
genr dyd=d(yd)

'The DF_GLS testing equation
smpl 2 !obs
equation eq2.ls dyd yd(-1)

'Storing the test statistic
tests(!k)=eq2.@tstats(1)

next

'Storing the 1%, 5% and 10% critical values
cvs(1)=@quantile(tests,0.01)
cvs(2)=@quantile(tests,0.05)
cvs(3)=@quantile(tests,0.1)

'Storing the ordered test statistics generated over the replications
vector stats=@sort(tests)