Good afternoon. I'm have a coefficient interpretation question. I have two tables that all have the same outcome variables. Table 1 has no county fixed effect and table 2 has a county fixed effect. The estimates are of the percentage of a census tract that identify as a specific demographic. Subsequently, they can only take on values between 0 and 100.

Table 1:

          Outcome_1   Outcome_2

Level_1   0.116       2.31

Level_2   2.39        1.29

Constant  0.282       0.290

Table 2:

          Outcome_1   Outcome_2
Level_1   -2.091      -0.210

Level_2   -3.502       -3.410

Constant  0.456       2.12

For table 1, I assume that the mean estimate for Level_1 and Outcome_1 is 0.116+0.282 and thus the mean estimate for Level_1 without a fixed effect is 0.398. For table 2, however, I add the county fixed effect and the mean estimate is negative. How should I interpret this?

  • $\begingroup$ Further information is needed please. The question seems to be along the lines of "why does the estimate for a fixed effect become negative after adding an additional covariate as a fixed effect?" Is that what you are asking/thinking ? There could be several reasons but without more it's hard to be confident, but first thing that springs out to me is that the added covariate is a confounder. If you could edit the question to show a summary of both models, that would help a lot. If working with R then use summary(my_model). $\endgroup$ Commented Dec 2, 2023 at 19:23


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.