test for heteroskedasticity with or without year dummies

[IBTISSEM ibtissem]

I want to ask a question about the dummies ( year dummies, industry dummies) and heteroskedasticity.
Especially, I want to know if we should include the year dummies when we test for heteroskedasticity.
Should we write this equation: reg Y X1X2 i.Year ? then hettest, rhs fstat. or Should we test without the year dummies? It means we only write reg Y X1X2; then hettest, rhs fstat.
I'm wondering if I should or not include year dummies because I find different results when I include year dummies. I find that I have heteroskedasticity when I don't include year dummies. But when I include year dummies, I find that my errors are homoscedastic. thx

[Carlo Lazzaro]

Ibtissem:
welcome to this forum.
There's an upward issue that, in my opinion, is worth mentioning: was -i.year- included among the predictors of your -regress- code? Does it make sense according to the data generating process? If the answer is yes, why kicking it out when postestimating?

[IBTISSEM ibtissem]

Dear professor,
Thank you for your reply. Yes, year dummies should be included in my regression. I do the testparm and I find that I have to include i.year. I do share your opinion. It is not normal to delete i.year when postestimating. so, I have to do reg Y X1X2 i.year then hettest, rhs fstat . I don't understand why I found homoscedasticity when I include year dummies and heteroskedasticity without year dummies. thx

[Carlo Lazzaro]

Ibitissem:
you can have a clearer picture of what's going on with and without -i.year- just visually inspecting via -rvfplot-.
As an aside, please call me Carlo, like all on (and many more off) this forum do. Thanks.

[IBTISSEM ibtissem]

Carlo,
Thank you for your reply.