Statalist The Stata Listserver

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

st: FE regression

From   "Ingo Brooks" <>
Subject   st: FE regression
Date   Thu, 15 Feb 2007 21:51:18 +0100

Dear all,

I have a large panel dataset for which I would like to estimate a
regression model of the following structure:

y_{i,t} = a + b * z_{i} + c * x_{i,t} + d * zx_{i,t} + e_{i,t}

where zx_{i,t} is an interaction between dummy variable z_{i} which is
one for individuals from an EMU country and zero otherwise and an
individual specific variable x_{i,t} which varies over time and across

Estimating the regression with OLS works fine: . reg y z x zx, cluster(id)

However, there is a risk of endogeneity in the model and I would
therefore like to estimate the regression by aid of fixed effects
regression. However, since there is no within-variation inherent in
variable z_{i}, variable z_{i} will be dropped from the estimation. In
contrast, the interaction term zx_{i,t} is time-varying and will
therefore not be dropped from the within regression.

What I wonder now is whether or not the interaction term may be
interpreted as it is interpreted in the case of pooled OLS (i.e. as if
variable z_{i} would be included in the regression) or if I have to
drop the interaction variable from the

. xtreg y x zx, fe cluster(id)

regression in order to obtain a meaningful regression model.

Thanks for your insights in advance.

*   For searches and help try:

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