Macro

[Tolulope Oladele]

Good day everyone. I am trying to generate 4 ols regression models. I have created globals for my control variables and for my outcome variables. Below is my code. I get an error message - unexpected end of the file. How do I resolve this?

global VAR = "white_nh black_nh other_nh hispanic age_u20 age2026 age2730 age3135 age4145 age4650 age5155 age5660 age6164 age65pl lessshs hsged hsbelow socol college under25k income25_50k income75_100k income100pl childless child_1 child_2 child_3 i.gendernew married under138fpl btwn138400 over400fpl"

global outcome = "eshi pvtng medicaid uninsured"

eststo clear

foreach y in $outcome {
foreach g in $VAR {
eststo: reg `y' i.notworking_cvd##i.expansion i.notworking_non##i.expansion $VAR

quietly estadd local NonExpMean = round(r(mean), 0.1) if workstatus==0
quietly estadd local ExpMean = round(r(mean), 0.1) if workstatus==0

}

[Joseph Coveney]

I don't know what eststo and estadd are, but you're mssing a close brace for one of your foreach loops. That would give rise to the error message that you got.

It seems as if it's a specification search. Is that what these two commands are for?

[Nick Cox]

You are asking there for 32 x 4 = 128 models not 4. I guess you want a single loop

foreach y in $outcome {

eststo: reg `y' i.notworking_cvd##i.expansion i.notworking_non##i.expansion $VAR

quietly estadd local NonExpMean = round(r(mean), 0.1) if workstatus==0
quietly estadd local ExpMean = round(r(mean), 0.1) if workstatus==0

}

That is, I guess that you don't want to loop over your control variables.

Joseph Coveney has a good point: eststo and estadd are community-contributed commands, and you're asked to explain where they come from -- in this case estout from SSC. See FAQ Advice #12. They are great commands -- which I never use because I never need the ritual display of elaborate results tables. So, I can't comment on whether your syntax will work otherwise.

But as explained in Section 4 of https://journals.sagepub.com/doi/pdf...867X1801800311 round() with second argument of 0.1 is a poor way to ask for a display format. If you want %2.1f as a display format, just ask for it.

A fact that will sometimes bite you is that Stata can't hold most multiples of 0.1 exactly -- because at machine level it works in binary. Most of the time Stata does its level best to protect you from this, but it is discernible.

. di %23.18f  round(1/10, 0.1)
   0.100000000000000006

If you want a deeper understanding of why it seems that Stata can't do arithmetic,

search precision

will show some resources.

[Tolulope Oladele]

Thanks so much Joseph and Nick. The code work on including the missing close brace.The estadd command can be found in http://repec.org/bocode/e/estout/estadd.html.
I want to add descriptive statistics of my regressors into my table. Thanks, Nick. I will look the article you shared on rounding.