Setting a condition when a variable must be larger than the maximum of other two

[Dario Maimone Ansaldo Patti]

Dear All,

I would like to set a conditional stattements in my gen command. Something like:

gen y=1 if x>max{k,z}

It might be trivial, but I cannot figure out how I can get something like that. Could you please offer me some suggestions?

Thanks in advance

Dario

[Nick Cox]

max(k, z) is fine as Stata syntax (parentheses not braces).

[Andrew Musau]

Almost exactly as you put it, except you need parentheses (so that the last term is the -max()- function). Missing values are evaluated greater than everything else, so you need care in formulating the expression.

sysuse auto, clear
keep in 1/10
keep rep78 headroom gear_ratio
gen wanted= rep78 > max(headroom, gear_ratio) & !missing(rep78) & !(missing(headroom) & missing(gear_ratio))

Res.:

. l, sep(0)

     +--------------------------------------+
     | rep78   headroom   gear_r~o   wanted |
     |--------------------------------------|
  1. |     3        2.5       3.58        0 |
  2. |     3        3.0       2.53        0 |
  3. |     .        3.0       3.08        0 |
  4. |     3        4.5       2.93        0 |
  5. |     4        4.0       2.41        0 |
  6. |     3        4.0       2.73        0 |
  7. |     .        3.0       2.87        0 |
  8. |     3        2.0       2.93        1 |
  9. |     3        3.5       2.93        0 |
 10. |     3        3.5       3.08        0 |
     +--------------------------------------+

[Dario Maimone Ansaldo Patti]

Thanks Nick Cox and Andrew Musau I just realized that it was a problem with brackets. THanks for confirming this.

[Joro Kolev]

The max function does not obey the usual Stata logic that missing are bigger than all non-missings. E.g.,

. dis max(1,.)
1

. dis max(.,.)
.

. dis max(1,2,3,.)
3

[Dario Maimone Ansaldo Patti]

Joro Kolev thanks for pointing this. I read something about. I did not care about this issue in this case, as i do not have any missing value in my dataset. But i will keep this information in mind for the future

[Nick Cox]

It's also Stata logic to ignore missings whenever possible. You would not think much of summarize if on being presented with 1,2,3,. it refused to answer on the grounds that a missing value makes the problem insoluble. Of course, there is software which behaves in precisely that manner.