MATA-looping through matrices & store strings in stata

[praveen wp]

Hi, I'm new to mata and I have two problems related to MATA. Can someone please help me to find solutions to these.

(1) Looping through matrices in MATA

mata

for (j=1;j<=3;j++)
    {
        X=(1,2\3,5)
        m_1=(1,1\1,1)
        m_2=(2,3\5,6)
        m_3=(5,3\8,9)
        X-m_j // here I need to refer to the jth matrix at each iteration. i.e. when j=1 m_ j should equal to m_1 and so on..

    }

end

The output will be

When j=1
1 2
+---------+
1 | 0 1 |
2 | 2 4 |
+---------+

When j=2
1 2
+-----------+
1 | -1 -1 |
2 | -2 -1 |
+-----------+

When j=3
1 2
+-----------+
1 | -4 -1 |
2 | -5 -4 |
+-----------+

(2) saving MATA string vector as variables of STATA

STATA code
gen a1=1
gen a2=1

This creates two variables named a1 and a2 in stata

I have a string vector in MATA. I need to do the same thing using MATA. (i.e. store the string names as variables)

mata
mat1="a1","a2"
// here I need the equivalent MATA code to generate the a1 and a2 variables
end

Thank you for your time

[Joseph Coveney]

Here is one way each to accomplish of your objectives.

.ÿversionÿ14.1

.ÿ
.ÿclearÿ*

.ÿsetÿmoreÿoff

.ÿquietlyÿsetÿobsÿ2

.ÿ
.ÿlocalÿlinesizeÿ`c(linesize)'

.ÿsetÿlinesizeÿ78

.ÿ
.ÿmata:
-------------------------------------------------ÿmataÿ(typeÿendÿtoÿexit)ÿ----
:ÿmataÿclear

:ÿmataÿsetÿmatastrictÿon

:ÿ
:ÿvoidÿfunctionÿpraveen1()ÿ{
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿmatrixÿX
>ÿÿÿÿÿÿÿÿÿXÿ=ÿ(1,ÿ2ÿ\ÿ3,ÿ5)
>ÿ
>ÿÿÿÿÿÿÿÿÿpointer(realÿmatrix)ÿvectorÿMatrices
>ÿ
>ÿÿÿÿÿÿÿÿÿMatricesÿ=ÿ&(1,ÿ1ÿ\ÿ1,ÿ1)
>ÿÿÿÿÿMatricesÿ=ÿ(Matrices,ÿ&(2,ÿ3ÿ\ÿ5,ÿ6))
>ÿÿÿÿÿMatricesÿ=ÿ(Matrices,ÿ&(5,ÿ3ÿ\ÿ8,ÿ9))
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿscalarÿj
>ÿÿÿÿÿÿÿÿÿforÿ(j=1;ÿj<=3;ÿj++)ÿ{
>ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿprintf("\n\nWhenÿjÿ=ÿ%1.0f\n",ÿj)
>ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿXÿ-ÿ*Matrices[j]
>ÿÿÿÿÿÿÿÿÿ}
>ÿ}

:ÿ
:ÿpraveen1()


Whenÿjÿ=ÿ1
ÿÿÿÿÿÿÿ1ÿÿÿ2
ÿÿÿÿ+---------+
ÿÿ1ÿ|ÿÿ0ÿÿÿ1ÿÿ|
ÿÿ2ÿ|ÿÿ2ÿÿÿ4ÿÿ|
ÿÿÿÿ+---------+


Whenÿjÿ=ÿ2
ÿÿÿÿÿÿÿÿ1ÿÿÿÿ2
ÿÿÿÿ+-----------+
ÿÿ1ÿ|ÿÿ-1ÿÿÿ-1ÿÿ|
ÿÿ2ÿ|ÿÿ-2ÿÿÿ-1ÿÿ|
ÿÿÿÿ+-----------+


Whenÿjÿ=ÿ3
ÿÿÿÿÿÿÿÿ1ÿÿÿÿ2
ÿÿÿÿ+-----------+
ÿÿ1ÿ|ÿÿ-4ÿÿÿ-1ÿÿ|
ÿÿ2ÿ|ÿÿ-5ÿÿÿ-4ÿÿ|
ÿÿÿÿ+-----------+

:ÿ
:ÿvoidÿfunctionÿpraveen2()ÿ{
>ÿ
>ÿÿÿÿÿÿÿÿÿstringÿvectorÿNewvarlist
>ÿÿÿÿÿÿÿÿÿNewvarlistÿ=ÿ("a1",ÿ"a2")
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿrowvectorÿNewvarindices
>ÿÿÿÿÿÿÿÿÿNewvarindicesÿ=ÿst_addvar("byte",ÿNewvarlist)
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿmatrixÿNewData
>ÿÿÿÿÿÿÿÿÿNewDataÿ=ÿJ(st_nobs(),ÿlength(Newvarlist),ÿ1)
>ÿ
>ÿÿÿÿÿÿÿÿÿst_store(.,ÿNewvarindices,ÿNewData)
>ÿ
>ÿÿÿÿÿÿÿÿÿstata("list,ÿnoobs")
>ÿ}

:ÿ
:ÿpraveen2()

ÿÿ+---------+
ÿÿ|ÿa1ÿÿÿa2ÿ|
ÿÿ|---------|
ÿÿ|ÿÿ1ÿÿÿÿ1ÿ|
ÿÿ|ÿÿ1ÿÿÿÿ1ÿ|
ÿÿ+---------+

:ÿ
:ÿend
------------------------------------------------------------------------------

.ÿ
.ÿsetÿlinesizeÿ`linesize'

.ÿ
.ÿexit

endÿofÿdo-file


.

[praveen wp]

Thank you very much Joseph. That solved my problem.

[praveen wp]

Just a quick question. Is there a easy way to achieve this than using pointers?
For example: in stata same code can be written

forval j = 1/3 {
    matrix X=(1,2\3,5)
    matrix m_1=(1,1\1,1)
    matrix m_2=(2,3\5,6)
    matrix m_3=(5,3\8,9)
    matrix A=X-m_`j'
    matrix list A
}

STATA uses ` ' to refer to the jth matrix. Thank you very much for your assistance.

[Jesse Wursten]

If the matrices are all the same dimension, you could combine the matrices and then refer to the correct columns of the matrix instead of using multiple matrices. If they are not of the same dimension, I think that's still possible, but you would have to create two vectors first which indicate the relevant first and last column. You'd then loop over these vectors to get the right column numbers in each iteration.

There's probably a more elegant solution, but that's all I can come up with.

[Joseph Coveney]

Just a quick question. Is there a easy way to achieve this than using pointers?
For example: in stata same code can be written

forval j = 1/3 {
matrix X=(1,2\3,5)
matrix m_1=(1,1\1,1)
matrix m_2=(2,3\5,6)
matrix m_3=(5,3\8,9)
matrix A=X-m_`j'
matrix list A
}

STATA uses ` ' to refer to the jth matrix. Thank you very much for your assistance.

You can achieve this without explicit use of pointers, for example, you can use an associative array (which would be using pointers behind the scenes). See below.

But nothing that I've seen in the documentation suggests that Mata has macro features that Stata has.

By the way, you're redefining the same matrices in each pass through the loop. Wouldn't it be better to assign the matrices just once before entering the loop?


.ÿversionÿ14.1

.ÿ
.ÿclearÿ*

.ÿsetÿmoreÿoff

.ÿ
.ÿlocalÿlinesizeÿ`c(linesize)'

.ÿsetÿlinesizeÿ78

.ÿ
.ÿmata:
-------------------------------------------------ÿmataÿ(typeÿendÿtoÿexit)ÿ----
:ÿmataÿsetÿmatastrictÿon

:ÿ
:ÿvoidÿfunctionÿpraveen1a()ÿ{
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿmatrixÿX
>ÿÿÿÿÿÿÿÿÿXÿ=ÿ(1,ÿ2ÿ\ÿ3,ÿ5)
>ÿ
>ÿÿÿÿÿÿÿÿÿclassÿAssociativeArrayÿscalarÿa
>ÿÿÿÿÿÿÿÿÿa.reinit("real")
>ÿ
>ÿÿÿÿÿÿÿÿÿa.put(1,ÿ(1,ÿ1ÿ\ÿ1,ÿ1)ÿ)
>ÿÿÿÿÿÿÿÿÿa.put(2,ÿ(2,ÿ3ÿ\ÿ5,ÿ6)ÿ)
>ÿÿÿÿÿÿÿÿÿa.put(3,ÿ(5,ÿ3ÿ\ÿ8,ÿ9)ÿ)
>ÿÿÿÿÿÿÿÿÿa.notfound(ÿJ(2,ÿ2,ÿ.)ÿ)
>ÿ
>ÿÿÿÿÿÿÿÿÿrealÿscalarÿj
>ÿÿÿÿÿÿÿÿÿforÿ(j=1;ÿj<=3;ÿj++)ÿ{
>ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿprintf("\n\nWhenÿjÿ=ÿ%1.0f\n",ÿj)
>ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿXÿ-ÿa.get(j)
>ÿÿÿÿÿÿÿÿÿ}
>ÿ}

:ÿ
:ÿpraveen1a()


Whenÿjÿ=ÿ1
ÿÿÿÿÿÿÿ1ÿÿÿ2
ÿÿÿÿ+---------+
ÿÿ1ÿ|ÿÿ0ÿÿÿ1ÿÿ|
ÿÿ2ÿ|ÿÿ2ÿÿÿ4ÿÿ|
ÿÿÿÿ+---------+


Whenÿjÿ=ÿ2
ÿÿÿÿÿÿÿÿ1ÿÿÿÿ2
ÿÿÿÿ+-----------+
ÿÿ1ÿ|ÿÿ-1ÿÿÿ-1ÿÿ|
ÿÿ2ÿ|ÿÿ-2ÿÿÿ-1ÿÿ|
ÿÿÿÿ+-----------+


Whenÿjÿ=ÿ3
ÿÿÿÿÿÿÿÿ1ÿÿÿÿ2
ÿÿÿÿ+-----------+
ÿÿ1ÿ|ÿÿ-4ÿÿÿ-1ÿÿ|
ÿÿ2ÿ|ÿÿ-5ÿÿÿ-4ÿÿ|
ÿÿÿÿ+-----------+

:ÿ
:ÿend
------------------------------------------------------------------------------

.ÿ
.ÿsetÿlinesizeÿ`linesize'

.ÿ
.ÿexit

endÿofÿdo-file


.

[praveen wp]

Thank you very much Jesse and Joseph. You have been really helpful..

[Christophe Kolodziejczyk]

The "use" of Stata Macros in Mata is explained in this Stata Journal article. In Mata Macros can't be used in the same way as in Stata, i.e for looping. Use pointers (or associative arrays as usggested by Joseph) to collect objects of different types or shapes. I think in this case pointers should be easy enough. For more complicated cases associative arrays would be a nice solution.