Insignificance with interaction term

[Tom Schmidt]

Hello,

some of my IV's in my regression model are Black (dummy), Union (Dummy) and School (Yrs of education).When I create an interaction term between Black/School, Black and the interaction term become insignificant. When I create the interaction term with Black/Union, Black remains significant. Can anyone explain this to me in simple words?

Thx
TS

[Eric de Souza]

If you expect an answer, please post the Stata output here -- and use codel delimiters to post it so that it is readable:consult the FAQ on how to do this:
https://www.statalist.org/forums/hel...ce/bbcode_code

[Marcos Almeida]

Interaction is an important issue when it comes to regression analysis. It seem the information you want is related to the is very principles. Perhaps a good start will be reading an introductory text on the matter.

[Richard Williams]

Once you add interactions, the meanings of the main effects change and significance tests may not mean much. For a discussion, see

https://www3.nd.edu/~rwilliam/stats2/l53.pdf

[Phil Bromiley]

"When I create an interaction term between Black/School, Black and the interaction term become insignificant." -- this means that the effect of Black is statistically insignificant when School=0. Essentially, instead of being estimated on the full sample, the main effect is estimated on the observations where the interacting variable is zero. "When I create the interaction term with Black/Union, Black remains significant." That means the effect of Black is statistically significant for the subsample of observations where union==0.