Years are dropped when running a fixed effect regression

[Laura Banken]

Dear all,

For my thesis, for which I have gathered panel data, I have to run a logit regression with year and industry fixed effects. I did this with the code:

'Logit y x1 x2 x3 i.sic i.fyear, vice(robust)'

My data has around 50.000 observations from 2010 through 2019.
However now I see, in the output, that the year dummy variable only starts at 2013, while my panel data starts in 2010. Does anyone know the reason why three years are skipped? In addition, for the last year of my data (2019), it is said that '2019 is committed because of collinearity'
Is this normal? I am a beginner in terms of STATA, so I have no idea...
Thank you!

[Carlo Lazzaro]

Laura:
I'm not clear why you stay with -logit- when -xtlogit- is available for panel data with binomial regressand.
Besides, -xtlogit- has an option for (conditionl) effect specification.
Without an example/excerpt of your data to be posted via -dataex- along with what you typed and what Stata gave you back (both these recommedations can be found in the FAQ), it is difficult to reply to your substantive queries more positively.
That said, what you reported can happen (as it has happened to you).
In finishing off, two asides:
- your -vice(robust)- is -vce(robust)-. Please note that, under -logit- -robust- takes heteroskedasticity into account, not serial correlation (in which you're presumably interestted, as your observations are not indeprendent);
- being a beginner in Stata is obviously a possible state of the world that affected anyone on this list at a given point in time (quoting one of Nick Cox 's comments that dates back to some years ago: "we are all beginners; some of us are only more experienced"); the issue is being familiar with the theory undelying panel data analysis for non-continuous regressands.