Problem setting base category in interaction term

[Felix Schaff]

Hi everyone,
I have a question about choosing the base level of a factor variable in an interaction term with main effects only when using xtreg. I presume I'm not the first person to run into this issue, but I cannot find any guidance in the Stata manual or old forum posts to help. The issue occurs when I am trying to set a certain base category in an interaction term. My choice of base category is ignored by Stata. Specifically, I try to set the second category of the variable x2 as base in an interaction of x1 with x2. However, in the output the fourth category of x2 is omitted, instead of the second category being the base, as specified (see below the relevant part of the output):

xtreg y c.x1#ib2.x2 i.year , fe robust allbaselevels 

                           |               Robust
                         y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
                   x2#c.x1 |
                        1  |   .8954887   .6693987     1.34   0.189    -.4584979    2.249475
                        2  |   .1291345   .6481673     0.20   0.843    -1.181908    1.440177
                        3  |  -.8268249   .4934912    -1.68   0.102    -1.825005    .1713553
                        4  |          0  (omitted)
                           |
                     year  |
                     1400  |          0  (base)
                     1425  |   .1343599   .1536929     0.87   0.387    -.1765132    .4452331
                     1450  |   .3906657   .2905451     1.34   0.187    -.1970173    .9783487
                     1475  |   1.199689   .4185224     2.87   0.007     .3531474     2.04623

...

Any tips would be greatly appreciated.

Thank you so much
Felix

[Rich Goldstein]

you have mis-specified your interaction and, as you can, see, x2 by itself is not shown at all; try

 
 xtreg y c.x1##ib2.x2 i.year , fe robust allbaselevels

that is, include the second # symbol

[Felix Schaff]

Thank you Rich! I deliberately wanted to have only the main effect, because x1 and x2 alone are absorbed by the unit-fixed effects and the time fixed effects. That is why I specified the interaction term only with one #. Do you think there is a way to make this work with only one #?
Best,
Felix

[Andrew Musau]

This question pops up here from time to time. See https://www.statalist.org/forums/for...nuous-variable

[Rich Goldstein]

re: #3 - I don't think I understand; using just one # means you ONLY get the interaction; if you don't want the interaction, don't use any #'s