puzzling out for num line of code

[natalia malancu]

hi guys!

I recently received a *do file dating back to 2010 and there's a line of code I'm struggling with:

for num 1/6: by xwaveid: replace hhfxid=hhfxid[X] if hhfxid==""

where xwaveid is the crosswave ID (7-digit) and hhfxid is the father's xwaveid (text, 7-digit)

Specifically I'm unable to figure out what the requested change is / how to look for the change made

grateful for your suggestions

[Clyde Schechter]

-for- is an obsolete command that was used before -foreach- was created.

If I remember how its syntax worked, a modern substitue for the code you show would be:

forvalues i = 1/6 {
    by xwaveid: replace hhfxid = hhfxid[`i'] if missing(hhfxid)
}

The effect of this code would be to replace missing values of the variable hhfxid with the first non-missing value among the first six observations of hhfxid (this is done separately within each group of observations corresponding to a single value of xwaveid). Does that seem sensible in your context?

[Nick Cox]

for in Stata went undocumented in Stata 7, if I recall correctly. It should still work, but it is not documented (not even undocumented).

Modern equivalent is

forval x = 1/6 { 
    by xwaveid: replace hhfxid=hhfxid[`x'] if hhfxid==""
}

In turn, the first hhfxid value in each wave is copied to all values for each identifier are missing.

If any are still missing, use the second value, and so on.

It doesn't check for inconsistencies in hhfxid between waves.

[natalia malancu]

thank you both for the modern equivalent and explanation. nick you are right: 'for' still works.

what I still dont get is the following:

if I keep just one wave of the data, the code generates no changes (no surprise)
if, however, I keep 2 waves, then for both `x'=1 and `x'=2 I get changes. shouldnt there be changes only for `x'=1? I mean if in wave 2 I have no ID value for the father, then it is copied from wave 1. what am I missing here?

[Nick Cox]

I can't see your data, but consider this dataset

. clear

. set obs 6
obs was 0, now 6

. gen father = "dad" in 2
(5 missing values generated)

. l, sep(0)

     +--------+
     | father |
     |--------|
  1. |        |
  2. |    dad |
  3. |        |
  4. |        |
  5. |        |
  6. |        |

Looking at the first observation won't change anything, but looking at the second will. After that no changes.

[natalia malancu]

the only explanation I can think of is:
the ID of the father is missing in wave 2 and not in wave 1 `x'=1
the ID of the father is missing in wave 1 and not in wave 2 `x'=2
would that make sense?

[natalia malancu]

guess I was typing as you were typing, and the answer is the same. thanks nick

[Carole J. Wilson]

I was a big fan of this construction and still keep these pages bookmarked:
http://www.stata.com/support/faqs/pr...-for-commands/
http://apps.eui.eu/Personal/Franklin...tyle%20for.pdf

[Nick Cox]

I think you are most of the way there. The code loops over the observations in each group. Run through the code on 6 observations with this pseudocode.

for i = 1 to 6
replace hhfxid with hhfxid[i] if any values of missing

Thus

replace hhfxid with hhfxid[1] if any values are missing
replace hhfxid with hhfxid[2] if any values are missing
replace hhfxid with hhfxid[3] if any values are missing
replace hhfxid with hhfxid[4] if any values are missing
replace hhfxid with hhfxid[5] if any values are missing
replace hhfxid with hhfxid[6] if any values are missing

So it finds the first non-missing of hhfxid and uses that to fill in any missings.

[Nick Cox]

Carole: No, no, no! (Insert friendly grin.) See also http://www.stata.com/support/faqs/pr...lems-with-for/ written in despair because of the many people who got bitten when they tried to combine this with other constructs. Terrible messes, terrible confusion.

[natalia malancu]

will do.
I dare 2 more questions:
a. how would I got about a before and after list of the data?
b. how would I perform a consistency check?

[Carole J. Wilson]

No, no worries! That was before I was comfortable with using macros. I keep them bookmarked because I do occasionally find old code that I need to translate. I've reformed my ways and find loops are much more intuitive!