>>
Statalist,
I am using the frontier command in Stata 8.2. I am running the following:
frontier lncur18 lnmedia lnsre lncais lncoord lnsupp lntotexp, uhet (f16perc un divorc)
and it works fine. However, when I try:
frontier lncur18 lnmedia lnsre lncais lncoord lnsupp lntotexp, uhet (f16perc un divorc,nocons)
I am getting a message:
invalid 'noconstant'
But when I try
frontier lncur18 lnmedia lnsre lncais lncoord lnsupp lntotexp, uhet (f16perc un divorc,nocons) uhet(X)
it works fine.
Has anybody encountered this problem before?
Many thanks
Dev
>>> [email protected] 07/28/05 4:15 am >>>
Glenn Hoetker wrote:
> I have panel data from 1986-2003. Logic and examining the data suggest
> that the dependent variable increased monotonically over time. Because I
> have some time-invariant variables, I am using random-effects (Hausman
> test indicates this is okay). What are the relative merits of either:
>
> a. Including a YEAR variable, which will range from 1985 to 2003 in
> value, or
> b. Including 7 dummies, one for each year (except the omitted year)
>
> Possibly complicating the issue (or not), I have two dummy variables,
> one of which is set to one for all observations after 1994 and the other
> set to one after 1998. I have 47 observations per year, so degrees of
> freedom are not critical.
> I have found little guidance in my standard texts. Thank you for any
> guidance &/or references.
Adding to Austin's (correct) comments, I think the issue of whether you
should have one time trend variable or several time dummies is very much a
trade-off, but I would strongly favour time dummies if the conditions are
favourable.
The best argument for having a _time-trend_ is that it's a much more
parsimonious indicator of temporal effects: if Y = gpdgrowth and X1 =
-0.252*time (p<.05), we can see at a stroke that as time progresses,
economic growth significantly declines. The trouble is, of course, that
this can be too crude an indicator: what if growth declined four times but
rose three times in our period of observation?
That, of course, is where _time dummies_ come in. With these, we can
perhaps model such idiosyncratic temporal effects much more intelligently.
The trouble here - as Glenn insinuated - is that if degrees of freedom are
at a premium, we may be forced to use a time-trend instead.
If DF is _not_ a problem, however, I would use time-dummies every time.
Certainly, anytime I have fitted fixed-effect models to my own data, I
have always found time dummies to be more satisfactory in explaining the
underlying dynamics of the model than using time-trends: the difference in
model-fit is often overwhelming. Moreover, I've often noticed changes in
sign, as well drastic shifts in significance, to the other variables when
fitting models using either approach.
I hope this helps.
CLIVE NICHOLAS |t: 0(044)7903 397793
Politics |e: [email protected]
Newcastle University |http://www.ncl.ac.uk/geps
Whereever you go and whatever you do, just remember this. No matter how
many like you, admire you, love you or adore you, the number of people
turning up to your funeral will be largely determined by local weather
conditions.
*
* 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/
This message has been checked for viruses but the contents of an attachment
may still contain software viruses, which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.
*
* 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/
