Using the foreach/forval commands for different categories of a categorical variable

[Siddharth Jamad]

I have a panel dataset with the following key variables: race, wealth, year. Race is a string variable with the following categories: black, Hispanic, white.

I wish to generate mean wealth for each race by year using the foreach/forval commands. I tried running the following code:

foreach i in black Hispanic white {
2. egen meanwealth_`i' = mean(wealth) if race==`i', by(year)
3. }

Stata returns the following error message: black not found r(111);

I then encoded the race variable, calling the new variable race1 and ran the following code:

forval i = 1/3 {
2. egen meanwealth_`i' = mean(wealth) if race1==`i', by(year)
3. }

The code worked perfectly but instead of having variables like meanwealth_black I now have variables like meanwealth_1.

How can I have my variables be named after their categories instead of numbers when using the foreach command?

Thanks in advance for any and all help.

[Carlo Lazzaro]

Siddharth:
welcome to this forum.
-forval- loops over numbers(not names).
Your code asks it for calculating -meanwealth_1- if [bla, bla; just to make the story shorter and not to be disrespectful] -race1==1- (and then 2 and eventually 3).
But the -men- function from -egen- is byable.
So you may want to try:

bysort race1 (year): egen meanwealth=mean(wealth)

The following toy-example might be helpful:

set obs 10
number of observations (_N) was 0, now 10

. g id=_n

. g y=runiform()*10

. g race="black" in 1/4

. replace race="white" in 5/7

. replace race="venusian" in 8/10

. encode race, g(num_race)


. label list
num_race:
           1 black
           2 venusian
           3 white

. forval i = 1/3 {
  2. sum y if num_race==`i'
  3. }

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           y |          4    1.952401    1.413651   .2855687   3.488717

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           y |          3    5.848207    2.775075    3.23368   8.759911

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           y |          3    4.302978    4.048035   .7110509   8.689333

. bysort race: egen wanted=mean(y)

. list

     +------------------------------------------------+
     | id          y       race   num_race     wanted |
     |------------------------------------------------|
  1. |  3   1.366463      black      black   1.952401 |
  2. |  4   .2855687      black      black   1.952401 |
  3. |  2   2.668857      black      black   1.952401 |
  4. |  1   3.488717      black      black   1.952401 |
  5. |  9   5.551032   venusian   venusian   5.848207 |
     |------------------------------------------------|
  6. |  8    3.23368   venusian   venusian   5.848207 |
  7. | 10   8.759911   venusian   venusian   5.848207 |
  8. |  6   3.508549      white      white   4.302978 |
  9. |  7   .7110509      white      white   4.302978 |
 10. |  5   8.689333      white      white   4.302978 |
     +------------------------------------------------+

.

[Nick Cox]

Note as a start that


0. The same results could be held in

egen meanwealth = mean(wealth, by(race1 year)

Otherwise you can get what you want in several ways. Here are some of them.

1. Apply separate to the results of #0.

2. rename your existing variables. One line

3. Loop over races.

foreach  i in black Hispanic white   {
    egen meanwealth_`i' = mean(wealth) if race == "`i'", by(year)
}

4. A parallel loop

tokenize "black Hispanic white"

forval i = 1/3 {
    egen meanwealth_``i'' = mean(wealth) if race1==`i', by(year)
}

5. Another parallel loop.

local names black Hispanic white

forval i = 1/3 {
       gettoken name names : names
       egen meanwealth_`name' = mean(wealth) if race1==`i', by(year)
}

[Carlo Lazzaro]

-men- reads -mean- in post #2. Thanks.

[Nick Cox]

Another typo: the first line of code in #3 should be

 
 egen meanwealth = mean(wealth), by(race1 year)

[Siddharth Jamad]

Perfect, the 3rd suggestion by Nick is what I was fussing over. Thanks a lot to you both.