Error in foreach loop with mata

[Bill Kani]

Hi all

My goal is to calculate an index using mata for multiple data files (200+ cities) in the working directory, "d:/cities". I have to use mata because some matrix operations are fast for my analysis. I calculate an index using mata, and save it to a temp file. The following code works for the one data file (one city):

************************************************** ************************************************** *****
clear all
infile str4 cityname jobs latitude longitude using "d:/cities/nyc.txt", clear

drop if jobs==0

mata:
jobs = st_data(.,"jobs")
n = length(jobs)
d = J(n,n,.) // Initiate the pair-wise distance matrix
for (i=1;i<=n;i++) {
for (j=1;j<=n;j++) {
d[i,j]=exp(-sqrt((longitude [i]-longitude [j])^2+(latitude [i]-latitude [j])^2))
}
}
index = cross(cross(jobs,d)',jobs)/sum(jobs)/sum(jobs)
st_numscalar("myindex", index)
end

local myindex=index
tempname memhold
postfile `memhold' str4 cityname myindex using "d:\cities\output\index_out.dta", replace
post `memhold' (cityname[1]) (`myindex')
postclose `memhold'
************************************************** ************************************************** *****

But if I use a foreach loop to repeat the above calculation for many cities, the code does not go through, but stops with an error message, -- break --

************************************************** ************************************************** *****
clear all
cd "d:/cities"
fs *.txt // Get the list of the file names
foreach f in `r(files)' {
infile str4 cityname jobs latitude longitude using "d:/cities/`f'", clear

mata:
jobs = st_data(.,"jobs")
n = length(jobs)
d = J(n,n,.) // Initiate the pair-wise distance matrix
for (i=1;i<=n;i++) {
for (j=1;j<=n;j++) {
d[i,j]=exp(-sqrt((longitude [i]-longitude [j])^2+(latitude [i]-latitude [j])^2))
}
}
index = cross(cross(jobs,d)',jobs)/sum(jobs)/sum(jobs) // This index is a scalar.
st_numscalar("myindex", index)
end

local myindex=index
tempname memhold
postfile `memhold' str4 cityname myindex using "d:\cities\output\index_out.dta", replace
post `memhold' (cityname[1]) (`myindex')
}
postclose `memhold'
************************************************** ************************************************** *****

I expected `myindex' values are stored in the "d:\cities\output\index_out.dta" for each city, but the program does not run. I would really appreciate it if someone could provide some help to tackle the problem.

Sincerely,

Bill K.

[Robert Picard]

You get the error because an end statement is interpreted as a break when encountered within a loop. The following illustrates the issue

foreach n in 1 2 {
    dis `n'
    end
}

Note that the display statement never executes. The solution has been pointed out by Maarten Buis on Statalist before:

http://www.stata.com/statalist/archi.../msg00393.html

[Christophe Kolodziejczyk]

I prefer Maarten's solution of writing a specific mata function for the task, so mata code lies in a different place and makes the do-file more readable. But if you want to avoid this solution you can use semi-colons as delimiters in your mata code. Note the use of the brackets to delimit the mata part. Here is an example

forval i = 1/2 {
  mata {
 
  for (i=1;i<=2;i++) {;
printf("i=%2.0f\n",i);
  };
 
  }
 
  }

[Bill Kani]

To Robert: Thanks for the link. I could get around my problem by creating a separate ado file.

To Christophe: Thanks for the tip. Maarten's solution looks elegant, but I equally like your solution. Yours worked nice for my project.

[Janys Ung]

Hi @Christophe Kolodziejczyk, I'm trying your solution with 2 statements in the Mata function but it does not work properly. It gives a message "code follows on the same line as close brace". Following is my code. Could you please help me to have a look?

forval n = 1/5{
    mata{
    
    for (n = 1; n<=5 ; n++){ ;
    b_M`n' = b_M`n'B5 - b_M`n'B1 ;
    z_M`n' = b_M`n'/(((se_M`n'B1^2/)+(se_M`n'B5^2/))^(1/2)) ;
    };
}
}

I would greatly appreciate your help!