Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Is xtmixed is the appropriate estimation method?

From   "Stephan Brunow" <>
To   <>
Subject   st: Is xtmixed is the appropriate estimation method?
Date   Fri, 4 May 2007 11:46:01 +0200

Dear STATAListers,

we want to estimate a Cobb-Douglas production function,
ln y_r=ln A+b ln K_r + c ln L_r

where K is capital and L is labor input. We have only observation of one
year for almost 200 different regions (labeled with _r). There is no panel
so far. However, we have detailed information of the labor force within each
region. I.e. we have information of single workers (_i) within each region
(_r). Therefore we have about 1 million of rows. We think that regional
parameters vary randomly because of the regional labor force. Hence, we


where age is the individual age of a person within each region. 
The data set looks as follows

Region  lny	lnK  	lnL	Personid	age
1       10 	5	3	1           23
1       10 	5     3	2           25
2       12	7     6	1           48
2       12  7    	6	2           57
2       12  7    	6	3           36

As you might see, the regional information is constant over all persons of
that region but the individual information varies.

If we define ageXlnL=age*lnL,

the xtmixed command is

xtmixed lny lnK lnL ageXlnl || region: lnL

However, this model does not work. We get positive values of the
log-restricted likelihood and the Hessian can not be computed.
Therefore we have chosen the emonly option which „only“ reports positive
values oft he log-restricted likelihood.

We know that we do not have an y_ri, we have only a y_r. Is this the
problem? Is xtmixed the right model at all?

Thank you,


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index