forval loop overwriting its previous iterations

[Topher McDougal]

I’m having a rather silly problem with a simple forval loop, such that the last cycle of the loop seems to be populating all previous local macros created therein. I am simply trying to record a series of r-squared values in a matrix over subsets of observations defined by a decile variable (here called var3). My code is:

forval i = 1/10 {
qui xtreg var1 var2 if var3==`i'
local rsquare`i' e(r2_o)
}

matrix define r2wide = (1,2,3,4,5,6,7,8,9,10\`rsquare1',`rsquare2',`rsqua re3',`rsquare4',`rsquare5',`rsquare6',`rsquare7',` rsquare8',`rsquare9',`rsquare10')
matrix r2 = r2wide'
matrix list r2

The resulting matrix is something like:

r2 [10,2]

	r1	r2
c1	1	0.245521
c2	2	0.245521
c3	3	0.245521
c4	4	0.245521
c5	5	0.245521
c6	6	0.245521
c7	7	0.245521
c8	8	0.245521
c9	9	0.245521
c10	10	0.245521

where, as you can see, all r-squared values have been replaced with the most recent value generated – even though if I insert a command line into the loop to display its output in media res (di rsquare`i' inserted just below local rsquare`i' e(r2_o)), I obtain a series of distinct results (but still the same homogenous matrix at the end).

If I run a similar forval loop without the regression and r-square extraction, there’s no problem:

forval i = 1/10 {
local rsquare`i' `i'
}

matrix define r2wide = (1,2,3,4,5,6,7,8,9,10 \ `rsquare1',`rsquare2',`rsquare3',`rsquare4',`rsqua re5',`rsquare6',`rsquare7',` rsquare8',`r2QtauS9',`rsquare10')
matrix r2 = r2wide'
matrix list r2

yields:

r2 [10,2]

	r1	r2
c1	1	1
c2	2	2
c3	3	3
c4	4	4
c5	5	5
c6	6	6
c7	7	7
c8	8	8
c9	9	9
c10	10	10

Many thanks.
Version: Stata 13.1 (up to date)

[Clyde Schechter]

The problem is in the first loop where you write

local rsquare`i' e(r2_o)

That command does not tell Stata to find the value of e(r2_o) and put it in local macro rsquare`i'. It tells Stata to put the string "e(r2_o)" in rsquare`i'. So each of the rsquare macaros contains the string "e(r2_o)". Then when you go to your -matrix define- command, Stata decides that you want to evaluate e(r2_o)--because how else could it make sense of that: you can't put strings in a Stata matrix. But at that point the only e(r2_o) hanging around is the final one. The solution is simple:

local rsquare`i' = e(r2_o)

[Topher McDougal]

Of course! Thanks for the quick reply.

[Anand dubey]

Hi! I am writing the following line of command to generate dummy's:
webuse nlswork
forval i= 68/88 {
gen dummy`i'= 0
}

forval i= 68/88 {
replace dummy`i'= 1 if union`i'>0
}

But since in my data set i dont have year 74 76 79 81 84 86. I could not run the forval loop, because by writing 68/88 stata is saying that 74 is missing.
Can anyone suggest what i can do if i don't have consecutive values in a loop?

[Rich Goldstein]

use the -foreach- command; see that last of the syntaxes at

h foreach

note that if you don't understand what Stata means by a "numlist" (which the foreach command calls for), you should see

h numlist