Create 2 new variables taking the "maximum" value and tag the variable with maximum value

[Fadzai Chikwava]

Dear All

Please assist to create

1. A variable which will take the maximum value from three variables (

sum_rough, sum_inst , sum_public)

for each record. Data is longitudinal and has repeated observations "sum..." for each time period. So basically the new variable is repeated for each record in the same time period same as the sum variables shown below.

2. Create another variable which will "tag" the name of the variable with the maximum obervations

Example data shown below:

input byte id float time float sum_rough float sum_inst float sum_public

1 1 3 2 2

1 1 3 2 2

1 2 4 6 8

1 2 4 6 8

2 1 0 2 3

2 2 3 6 8

2 3 4 5 1

[Ali Atia]

egen maxvar = rowmax(sum_*)
gen whichvar = ""
foreach var of varlist sum_*{
    replace whichvar = "`var'" if maxvar == `var'
}

[Nick Cox]

Both the question and @Ali Atia's helpful reply leave open what to do if two or more values tie for maximum.

The maximum is the maximum, regardless, but which variable holds the maximum is more complicated.

In practice, that may not bite. For discussion of what to do if it does, see https://www.stata-journal.com/articl...ticle=pr0046_1

[Fadzai Chikwava]

Thanks for the code-very useful

Is there an easier way / syntax to handle situations where there is two or more values which tie for maximum.

[Fadzai Chikwava]

My second question which may help overcome the ties is: Is there a syntax where I can pick up the variable the "whichvar"- where the variable is highest is rank i.e. the numbers in the original variables represent the ranking order:

e.g. original variables and ties are in the following:
sum_1rough
sum_2inst
sum_3public

then if the maxim value is a tie on the three I would select "sum_1rough" which is highest in the ranking order.

[Nick Cox]

This is covered by the paper cited in #3. What suffices here is to select the first maximum you see and ignore the others. Tweaking the nice code of @Ali Atia

egen maxvar = rowmax(sum_*)

gen whichvar = ""

foreach sffx in 1rough 2inst 3public {
    replace whichvar = "`sffx'" if maxvar == sum_`sffx' & whichvar == "" 
}

[Fadzai Chikwava]

Thank you for this