Re: st: FW: Lambda in Frontier when using uhet

[Scott Merryman <scott.merryman@gmail.com>]

On Sun, May 6, 2012 at 8:13 PM, Huerta, Tim <tim.huerta@ttu.edu> wrote:
> According to Rosko - In developing a preferred model the following
> decisions had
> to be made: (1) should OLS or SFA be used; (2) what should be the
> structural form of the cost function; (3) what theoretical distribution
> should the composed error follow; and (4) should inefficiency-effects
> variables be included?
>
> It would seem that when someone is performing an SFE analysis in Stata and
> variables are loaded into the uhet, the diagnostic data to answer the
> first question isn't provided. In contrast when the uhet is not included,
> sigma_v, sigma_u etc. are provided - allowing for the researcher to answer
> q1.
>
> Can someone tell me why including variables in the uhet causes the
> information to disappear?

Rosko?

The information doesn't disappear, but if u is heteroskedastic then
there is no lambda, or rather there is a lambda for each observation
since sigma_u varies by observation.

Before seeing if the frontier model reduces to OLS, test the
homoskedasticity restriction that var1-var5 = 0.

Scott
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/