eclass store results

[Santiago_C]

Dear stata users,

I would like to store a matrix and the number of observations after my eclass program. In this simple example the output is a 6x6 matrix but of course it can contain n elements of the varlist. I am confused because some columns are varying strings which should be stored as macros or colnames? Notice that strings have different categories depending on the other (numeric) variables.

variable	sd	rh0(min)	cy	rh0(max)	ph
x1	1.145	0.950	string1	0.950	string1
x2	1.155	0.651	string2	0.660	string2
x3	1.035	0.25	string3	0.254	string3
x4	0.542	0.550	string4	0.552	string4
x5	0.663	0.247	string5	0.250	string5

Many thanks
Santiago C

[Maarten Buis]

There is no string matrix in Stata. So forget about that. Even in Mata you cannot have a mixed type matrix (*).

To find a meaningful solution to your problem we need to know what your problem is. Right now we have a classic XY problem: https://en.wikipedia.org/wiki/XY_problem . So tell us what you want your program to do, what are those columns, how do you expect the user to use what your program returns.

(*) Not quite true, but you don't want to go that route. Really trust me on that.

[Santiago_C]

Many thanks for the reponse. Indeed you are right, what I actually did is a program to report some variables with different conditions in terms of correlations. Below I have the last lines before the end
First I put a I _newline as text "variable" _column(1) as text "sd", then I make another newline with a different string name but then I express column(1) as result %5.3f `sd'.
For the other conditions I use the same structure with different columns. Results are clean, the only issue is that I want to store the matrix and the total number of observations to then export the results.

[Maarten Buis]

I still don't understand what you want to do. This is what I do understand:

• You have written a program
• It produces some output containing numbers and strings
• You want to return that output in a matrix

The last step is impossible, because it is a mix of numbers and strings, as I told you before. So forget about that. It cannot be done.

There are other things that can be done, but to decide what can be done we need to know what you want to do. So to be able to find a solution, I need you to answer, in detail, all these questions: want is your program supposed to do, what are those columns in the output, how do you expect the user to use what your program returns.

[Santiago_C]

Thanks again for responding.
Of course the last step cannot be done. I will keep the non-string variables or make more lines of coding to treat it in a different way.
Regards

[FernandoRios]

Hi Santiago
assuming you already know how to store the information you need either in locals, or matrices, here is a good trick.

program adde, eclass
ereturn `0'
end

So, after running your regression, you can add more info to e() using this command
for example:

sysuse auto,clear
reg price mpg headroom trunk weight length foreign
tabstat price mpg headroom trunk weight length foreign, save
matrix sstat=r(StatTotal)
adde matrix sum_stats = sstat
adde local model_des "This regression tries to explain prices as function of car characteristics"
ereturn list