The use of slash in Stata

[Attaullah Shah]

While I know that the following command will sum ri where the variable pic is equal to USA and spread the sum to all observation whether or not pic is equal to USA. However, I am not sure what specifically the slash does that the if condition cannot. also, what other useful uses of slash are availabe in Stata.

egen GDPref = total( ri / ( pic == "USA"))

[Nick Cox]

Attaullah omits the context of http://www.statalist.org/forums/foru...-a-local-macro

Here, as usual outside filenames, slash just means divide: no more, no less.

But in Stata dividing by 0 in particular means a missing result, sometimes bad news, but not necessarily in contexts where missings are ignored and that is precisely what you want.

Here the total() function just ignores missings and in the particular example that ensures that the values for the USA are spread to other values in the same year, thus providing a reference level.

More generally, the construct is useful mainly because of the way it works for true and false results for the logical comparison, in this case

pic == "USA"

False as a result for that comparison followed through means 'divide by 0' and people's mathematical education usually imparts a strong sense of sin, impropriety or danger to that division. But the effect can be what you want.

Studying the following example will show that what looks like the equivalent if condition has the side-effect that missings are generated whenever that condition is not true. But doing it as above means that non-missing values are returned routinely.

The example below uses groups from SSC.

Note that

. egen invest45 = total(cond(year == 1945, invest, .)), by(company)

has precisely the same effect.

Much more at http://www.stata-journal.com/sjpdf.h...iclenum=dm0055

. webuse grunfeld

. su year

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
        year |        200      1944.5    5.780751       1935       1954

. egen invest45 = total(invest / (year == 1945)), by(company)

. egen invest45_2 = total(invest) if year == 1945 , by(company)
(190 missing values generated)

. groups company invest45

  +--------------------------------------+
  | company   invest45   Freq.   Percent |
  |--------------------------------------|
  |       1      561.2      20     10.00 |
  |       2      258.7      20     10.00 |
  |       3       93.6      20     10.00 |
  |       4      88.78      20     10.00 |
  |       5      63.21      20     10.00 |
  |--------------------------------------|
  |       6      39.03      20     10.00 |
  |       7      44.12      20     10.00 |
  |       8      39.27      20     10.00 |
  |       9      52.32      20     10.00 |
  |      10       1.36      20     10.00 |
  +--------------------------------------+


. groups company invest45_2, missing sepby(company)

  +--------------------------------------+
  | company   invest~2   Freq.   Percent |
  |--------------------------------------|
  |       1      561.2       1      0.50 |
  |       1          .      19      9.50 |
  |--------------------------------------|
  |       2      258.7       1      0.50 |
  |       2          .      19      9.50 |
  |--------------------------------------|
  |       3       93.6       1      0.50 |
  |       3          .      19      9.50 |
  |--------------------------------------|
  |       4      88.78       1      0.50 |
  |       4          .      19      9.50 |
  |--------------------------------------|
  |       5      63.21       1      0.50 |
  |       5          .      19      9.50 |
  |--------------------------------------|
  |       6      39.03       1      0.50 |
  |       6          .      19      9.50 |
  |--------------------------------------|
  |       7      44.12       1      0.50 |
  |       7          .      19      9.50 |
  |--------------------------------------|
  |       8      39.27       1      0.50 |
  |       8          .      19      9.50 |
  |--------------------------------------|
  |       9      52.32       1      0.50 |
  |       9          .      19      9.50 |
  |--------------------------------------|
  |      10       1.36       1      0.50 |
  |      10          .      19      9.50 |
  +--------------------------------------+

[Attaullah Shah]

Thank you Nick, great explanation