[rezza cav]

However, I do not want certain variables to be prioritized and conditions to be applied sequentially.

I want the conditions to apply to all variables individually.

Is this possible at all? Maybe by first replacing the different values within a group variable by variable, and only then combining the observations?

[Andrew Musau]

You will need to determine the max of all other variables first, e.g., by generating a variable.

local all_others "x1 x2 x3 x4"
egen xmax = rowmax(`all_others')
by entity year firm, sort: gen freq = _N
gsort entity year -freq -xmax
by entity year: keep if _n == 1

If this does not answer your question, provide a data example as you did in #10 showing an initial and final dataset.

[rezza cav]

I want to obtain only one observation per combination of entity and year.

For the variable "firm", I want the most frequent value within the combination of entity and year to be chosen (to represent the former group of observations in the now single observation). If there is a tie, choose the first value within the group.
For the variable "turno", I want the most frequent value, if there is a tie, choose the highest value (max).
For the variable "dom", I want the most frequent value, if there is a tie, choose value 1 over value 0 (in this case max).

For all remaining variables (not shown in the example), choose the first value within the gruop (firstnm).

This is my original data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte entity int year str1 firm turno dom
1 2010 "A" 15 0
1 2010 "B" 8 0
1 2010 "B" 12 0
2 2011 "A" 25 0
2 2011 "A" 8 1
2 2011 "A" 7 0
3 2010 "A" 12 1
3 2010 "B" 5 1
3 2010 "B" 12 1
3 2010 "A" 8 1
end

sort entity year
list, sepby(entity year)

and I want to achieve this result:

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte entity int year str1 firm turno dom
1 2010 "B" 15 0
2 2011 "A" 25 1
3 2010 "A" 12 1
end

sort
entity year
list, sepby(entity year)