st: Re: Inferences from Quadchk applied to xttobit

["jyotsna puri" <pjyotsna@hotmail.com>]

Dear Listers,

Apologies for two queries in the same day - but they are different and I am
am a little stuck:

Background: I have a panel dataset  (10 year, two yearly dataset, 631
villages over 23 districts), with many corner solutions for my
dependent variable (area) i.e. zeroes. I use xttobit and then use quadchk
to see if my specifications are stable.

My Questions:
1. Can I infer anything from the 'change' in relative difference or
even the 'extent' of instability? i.e that my intra-panel correlation
is increasing too much for the polynomial apprimation in the gauss-hermite
procedure to be appropriate?

2. I almost always get a very unstable sigma_u ~=0.71, when
I increase the quadrature points from 12 to 20. Does
this mean I must reject the entire specification? My other
coefficients are relatively stable.

3. What does missing values for standard errors in xttobit imply?
In the following Lsr1tara is the log of area, popn and popnsq
are population variables, trdmin is the travel time to district
and agwsc2 is a dummy for water availability. If I alternatively
specify a model where an additional dummy variable for
land security is added, I get all the standard errors.

. xttobit Lsr1tara year popn popnsq trdmin trdminsq agwsc2 , i(ampname) 
ll(0)

Obtaining starting values for full model:
[output omitted]
Iteration 3:   log likelihood =   -5132.12

Fitting full model:
[output omitted]
Iteration 5:   log likelihood = -4323.7663

Random-effects tobit regression                 Number of obs      =      
2436
Group variable (i): ampname                     Number of groups   =        
23

Random effects u_i ~ Gaussian                   Obs per group: min =        
23
                                                              avg =     
105.9
                                                              max =       
327

                                                              Wald chi2(2)  
     =         .
Log likelihood  = -4323.7663                    Prob > chi2        =         
.

------------------------------------------------------------------------------
   Lsr1tara |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
       year |  -.0753494          .        .       .            .           
.
       popn |   .0038105          .        .       .            .           
.
     popnsq |  -1.01e-06   1.20e-07    -8.40   0.000    -1.25e-06   
-7.75e-07
     trdmin |  -.0090877          .        .       .            .           
.
   trdminsq |   5.81e-06   1.20e-06     4.83   0.000     3.45e-06    
8.16e-06
     agwsc2 |   1.147029          .        .       .            .           
.
      _cons |   7.900947          .        .       .            .           
.
-------------+----------------------------------------------------------------
   /sigma_u |   1.218991          .        .       .            .           
.
   /sigma_e |   2.944445   .0639365    46.05   0.000     2.819132    
3.069758
-------------+----------------------------------------------------------------
        rho |   .1463158          .                             .           
.
------------------------------------------------------------------------------

 Observation summary:      1417     uncensored observations
                           1019  left-censored observations
                              0 right-censored observations





Jyotsna (Jo) Puri
Cell and Voicemail: 201-805-1690
pjyotsna@hotmail.com
Doctoral Candidate, Department of Agriculture and Resource Economics,
University of Maryland, College Park, MD
I ran the Marine Corp Marathon to help educate.
http://www.ashanet.org/nycnj/events/2004/hoh/runners/jyotsna.html


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/