ereturn post a vector as a pseudovariable like e(sample)

[Yugen Chen]

Dear Statalist,

I would like to define an e-class program but have no idea how to -ereturn- a pseudovariable (xx in my example) like the -e(sample)-. Such a pseudovariable would be used in the following post-estimation command so I tried to use -eret post-. But I couldn't find how to do it from the help file for -eret-.

Any suggestions are appreciated!

Kind regards,
Yugen

// x = [1,2,3,4,5,6]
// y = [1,1,1,1,1,1]

clear all

program define myprograme 

args y x 

marksample touse 

preserve

quietly: keep if `touse'

qui: gen xx = x-3
qui: replace xx = 0 if x<0

/* some operations on y and x and xx */

restore 

// ereturn xx and get a psuedovariable like e(sample) 

end

[Jared Greathouse]

I don't think it's possible to ereturn a variable......... right?

[Joro Kolev]

I have never heard the terminology "pseudovariable" and I do not quite understand the question. e(sample) is a function I think.

What I can say is that what you are trying to do (which I do not completely understand) probably cannot be done your way with -preserve/restore-.

At this thread here #3 https://www.statalist.org/forums/for...ile-is-so-fast
Daniel made the good point that when you do -preserve/restore- you lose track of your sample.

[Yugen Chen]

Dear Jared and Joro, thanks for your replies!

1. "pseudovariable" is from the book An introduction to Stata Programming (2nd Ed) by Christopher Baum. Page 94 -e(sample)- is called a Stata "pseudovariable". Footnote 7 writes " Although -ereturn list- output describe -e(sample)- as a function, it is perhaps better considered a variable, albeit one that does not appear in the dataset". Since I don't want to make any changes to my dataset, I need to generate a "pseudovariable" like -e(sample)- which is actually an ereturn function from Stata documentation.

2. The program is used to calculate a local linear estimation around a specified point. I would like to store the linear weights say w_i on each observation (x_i, y_i). Then the estimator is given by \sum_i (w_i*y_i) where w_i is a function of {x_1 .. x_n} and will be used in one post-estimation command for plotting.