Dear DPD for Ox users,
I am a new DPD for user. I would like to ask two small questions concerning
robust heteroskedasticity in static panel data estimation. See for example
pp. 15-19 in the guidebook "Panel Data estimation using DPD for Ox,
December 23, 2002" by Doornik, Arellano and Bond.
1) For example, in the grunest1.ox -example, it is estimated the within
group estimates. By the command: dpd.SetOptions(TRUE), it is possible to
calculate robust standard erros. Is this an asymptotic variance estimate
for fixed T and large N suggested by Arellano 1987? Thus, you first
estimate the WG -residuals and then use these in calculating the asymptotic
variance?
2) For example, let's think the following model:
dpd.Select(Y_VAR, {"y", 0, 0}); // formulate model
dpd.Select(X_VAR, {"x1", 0, 0, "x2", 0, 0, "x3", 3, 3}) // // explanatory
variables
dpd.SetDummies(D_TIME + D_CONSTANT); // set dummies 1992-2002
Thus, you would like to estimate something with three explanatory variables
and the set of dummies, when one explanatory variable (actually a dummy
variable) is lagged three periods. After runnig the Ox, the output looks
something like:
---- 1-step estimation using DPD ----
Coefficient Std.Error t-value t-prob
x1 0.538500 0.06721 8.01 0.000
x2 0.201362 0.04224 4.77 0.000
x3(-3) -0.00186175 0.04838 -0.0385 0.969
T1996 0.0618537 0.03086 2.00 0.045
T1997 0.108043 0.03218 3.36 0.001
T1998 0.0974383 0.03417 2.85 0.004
T1999 0.134114 0.04279 3.13 0.002
T2000 0.194467 0.03876 5.02 0.000
T2001 0.195641 0.03952 4.95 0.000
T2002 0.214206 0.04332 4.94 0.000
My question is, should the Ox also use the T1993, T1994, T1995 dummies in
estimation? When I checked this in the Stata, it seems to maintain the
dummies in estimation.
If someone can help, I would appreciate it lot.
Best regards,
Mikko
PhD student
|