Interpretation OLS residual

[enneite enyx]

Hi,

I don't know how interpret the OLS of residuals. There is my code :

*model 4 CD
regress roi_closing_dom tx_var_dom trj_closing champion_dom i.paysnum i.id_liguenum b9.num_mois i.num_semaine b6.jour_semaine
predict cdd, residual
gen logcdd2 = log(cdd^2)
regress logcdd2 tx_var_dom trj_closing champion_dom i.paysnum i.id_liguenum b9.num_mois i.num_semaine b6.jour_semaine
predict zg

And here, there is my regression (where y = logcdd2). What are important things to interpret ? Thanks for you help !

[Carlo Lazzaro]

Enneite:
I've already replied on the same regression (that still looks poorly specified) at https://www.statalist.org/forums/for...variable-is-in.
That said, I'm not clear with your doubt about residual interpretation (that, in addition. you do not onclude in your post).
OLS residual distribution is usually inspected to detect heteroskedasticity and go -robust-.

[enneite enyx]

Hi Carlo,

Regarding your last sentences, OLS residuals is used to detect heteroscedasticty. So, in this OLS, how can I detect the heteroscédasticity? When coefficient are significant ? How can I interpet for example the coefficient of tx_var_dom ?
Thanks to you

[Carlo Lazzaro]

Enneite:
-estat hettest- is what you're looking for to check for heteroskedasticity.
As you have a log linear model, each 1-unit increase in -tx_var_dom- increases the regressand of about 0.96%:

. di (exp(.0095277)-1)*100
.9573233

[Dirk Enzmann]

You should consult a textbook. Some material you can find online, for example: https://stats.idre.ucla.edu/stata/we...n-diagnostics/