Dear researchers,
I might have asked this question before, but I didn't get an adequate answer, I might have asked the question in an unclear way.

I have an unbalanced dataset for the period extending from 2000-2012. Firms started to adopt a specific policy voluntarily in 2005 knowing that the adoption is not simultaneous. I mean firms adopted the policy in different years. For instance, a group of firms adopted in 2005, while others adopted in 2006, and so on. The main aim is to study what factor drives firms to implement these policies. I have applied the discrete hazard proportional odd regression “Multi-period logistic regression”. So, I had to delete all observations after the implementation. The dependent variable is the policy coded zero for all firm-year observations before the adoption and coded 1 in the year of adoption, where all observations after the adoption have been deleted. Also, I have built a calendar time variables. Then, I have applied the following code:

Code:
 encode Companyname , gen (COMPID)
global id COMPID
lab var TIME "SPELL YEAR"
lab var Event"Binary depvar for discrete hazard model"
ta TIME, ge (d)
ds d*
ge e1 = TIME <=4
ge e2= TIME >=5 & TIME <=7
ge e3= TIME >=8 & TIME <=10
logit Event Var1 Var2 Var3 e2 e3, nolog cluster (COMPID)
Then, I have received a comment from a journal to apply also the logistic regression, so I had to return all observations after the adoption for each firm, and I have applied the following code:
Code:
xtlogit Dep X1 X2 X3  i.Year, re 
 xtlogit Dep X1 X2 X3  i.Year, fe
I have found that all years have been dropped from the output tables in the STATA, so in my case I should use the above two commands without i.year?????

Many thanks in advance.