(Level model vs. Log model) Getting coefficients of opposite sign

[Chaeyeon Song]

Hello,

For my research paper, I am using two different models (level-level model and log-log model). If I estimate two different models, I get a negative coefficient on the variable of interest from the level-level model and a positive coefficient on the variable of interest from the log-log model. I personally find this result not intuitive. Can anyone tell me what these results suggest?

I am not referring to any dataset or regression results since I think the explanation here is sufficient.

[Joao Santos Silva]

Dear Chaeyeon Song,

It it difficult to help you without knowing more about what you are doing, but please try a Poisson regression where you regress levels on logs.

Best wishes,

Joao