Calculate % coefficient effect after regression

[John Economou]

Hey everybody,
I have a regression:

xtreg IMD ulc GDPpc capitalform mor sen  pop pat fd van_index i.year , fe ro

and I would like to calculate the % ulc effect which is the ratio of the coefficient divided by the linear prediction (and multiplied by 100), however I don't know how to calculate the "linear prediction" after the FE regression (the coefficient of ulc in my model is -0.018).

Thank you in advance!
Stay safe!
Sincerely,
John Economou.

[Clyde Schechter]

You need to clarify what you mean by the model's linear prediction. Do you mean for it to include the absorbed panel effect, or not? Either way, you would use the -predict- command after -xtreg-. If you want the panel effect included, it's -predict linear_effect, xbu-. If not, it's -predict linear_effect, xb-.See -help xtreg postestimation- and click on the blue link called predict for more information.

[John Economou]

Clyde, thanks a lot for your response!
However, I think that I want to calculate something else.
Specifically, the authors of this paper (https://www.sciencedirect.com/scienc...13001748#t0010) calculate the 'linear prediction' of the model, which is a number (e.g. 0.0159), not a vector (see Table 2).
Simirarly here http://ftp.iza.org/dp10633.pdf (see Table 4, etc.).
I want something like those...

Thanks again!

[Clyde Schechter]

The first of those papers is behind a paywall. The second is freely available, and uses the term linear prediction freely but never provides a definition or formula. I'm afraid I don't know what they are referring to. The only meaning I have ever known for the term is the one calculated by -predict-, but clearly that is not what you need. I'm sorry.

[Andrew Musau]

Specifically, the authors of this paper (https://www.sciencedirect.com/scienc...13001748#t0010) calculate the 'linear prediction' of the model, which is a number (e.g. 0.0159), not a vector (see Table 2).

So here they estimate a linear probability model, i.e., linear regression with a binary dependent variable (coded 0/1). The prediction is the average predicted dependent variable which is what margins gives you by default.

sysuse auto, clear
regress foreign mpg weight disp gear
predict foreignhat, xb
sum foreignhat
margins

Res.:

. predict foreignhat, xb

.
. sum foreignhat

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
  foreignhat |         74    .2972973     .334918  -.4234509    .941747

.
. margins

Predictive margins                              Number of obs     =         74
Model VCE    : OLS

Expression   : Linear prediction, predict()

------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   .2972973   .0377362     7.88   0.000     .2220157    .3725789
------------------------------------------------------------------------------

[John Economou]

Thank you Andrew!