How to generate different matrices based on different factors and different factors levels

[fu gang]

mata
D = (1,1,1 \ 2,1,1 \ 3,1,2 \ 4,1,2 \ 5,1,3 \ 6,1,3 \ 7,2,1 \ 8,2,1 \ 9,2,2 \ 10,2,2 \ 11,2,3 \ 12,2,3)
// two factor variables,the second column has two levels(1 2), the thrird column has three levels(1 2 3) ,then I will do loop on the two factor variables
for (i=1; i<=2; i++){
for (j=1; j<=3; j++) {
Dij=st_select(D,ij) \\ gen Dij matrix when the second column select i(1 or 2) and the third column select(1 or 2 or 3), then we get six matrixes:D11 D12 D13 D21 D22 D23
colnames = "_i_j" \\ rename the colnames of matrixes Dij
Dij = colnonmissing(Dij)' \\ then perform the same operation on the generated matrix
}
}
end

i don't know how to write the two rows of program in red, can you help? thank you very much

[Niels Henrik Bruun]

Maybe this help:

sysuse auto, clear
mata: D = st_data(., "i.rep78")
mata: D[.,1] :* D[.,2]

[fu gang]

thank you for your reply,thank you very much. using dummy variables , I have two Influencing factors，not one, and there are multilevels in each factor. I still can't do the loop and split the data, can you ，Can you explain further？

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte(obs a b) double v1
 1 1 1 1.2
 2 2 2 1.8
 3 1 3 1.6
 4 2 1 1.5
 5 1 2 1.9
 6 2 3   1
 7 1 1 1.1
 8 2 2 1.5
 9 1 3 1.6
10 2 1 1.7
11 1 2 1.5
12 2 3 1.6
end

a and b are two factor variables,v1 is the valuse for computing. We will select the data when a= (1 2) and b =(1 2 3),

[fu gang]

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte(obs a b) double v1
 1 1 1 1.2
 2 2 2 1.8
 3 1 3 1.6
 4 2 1 1.5
 5 1 2 1.9
 6 2 3   1
 7 1 1 1.1
 8 2 2 1.5
 9 1 3 1.6
10 2 1 1.7
11 1 2 1.5
12 2 3 1.6
end

/* Do one loop first */
mata
D = st_data(.,.)
for (i=1; i<=2; i++){
D1=st_select(D,D[.,2]=1,.) /* mistake here wrong expression of rows*/
D1
}
end


D1=st_select(D,1\3\5\7\9\11,.) when i =1, i want to get the rows of D[.,2] (variable a ) ==1 ,
D1 No results displayed， I don't get the target matrix, maybe I should use st_data()

mata
D = st_data(.,.)
for (i=1; i<=2; i++){
D1=st_data(v,.) /* I don't know how to express the target matrix rows, when i = 1, v shoud be the rows of matrix when a = 1*/
D1
}

D1=st_data(D,1\3\5\7\9\11,.) when i use st_data(), D1 is the target matrix, so st_data() will work.

When there are two or multiple cycles, how to take the intersection, that is, satisfy i=1 & j=1, That is, how to express the rows number of a=i and b=j through the expression of i and j

[fu gang]

mata:
mata clear
mata: D = st_data(., "i.a")
D
A = D[.,1] :* D[.,2]
A

The result is a matrix (0:0:0:0:0:0:0:0:0:0:0:0）but it is not what i want.

[Niels Henrik Bruun]

I misunderstood what you wanted.
Can you show input and output? And explain what you want?

[fu gang]

thank you very much , i wil show it right now

[fu gang]

I have a data, including the sequence obs, categorical variables a (1, 2) ，and b (1, 2, 3), and some other variables that need statistics v1, v2, v3...vn, omitted as v1
Now I want to split the data by a and b use mata. split thd data into 6 small data, D11, D12, D13, D21, D22, D23, when a=1 and b = 1, i'll get D11,and so on. and then calculate the number of non-missing values in the data, finally summarize the non-missing values of the six small data into one Table, If possible, calculate the sum of rows and columns. I can use stata to complete this calculation，but I am learning mata and want to use mata's loop to achieve

[fu gang]

Calculate the number of non-missing values in each small data Dij, and then merge the results into one table,calculate the sum of rows and columns, and finally my result is the number of non-missing values of v1, v2...vn variables separately，the target result is a table as follows:

count of non-missing values
D11 D12 D13 D21 D22 D23 Total
V1
V2
V3
...
Vn
Total

[Niels Henrik Bruun]

I think I would solve this in Stata instead of Mata

[fu gang]

thank you ,i have solved this in Stata for several methods
local v1 1 2
local v2 1 2 3

foreach i of local v1{
foreach j of local v2{
keep if a == `i'
keep if b == `j'
tabstat ……
}
}

maybe forvalues is better

[fu gang]

or
tabstat varlist if a == `i' & b ==`j', stat(count) col(stat) save
or
tabstat varlist if a == `i' , by(`j') stat(count) col(stat) save

[fu gang]

Maybe you have several better and more concise methods, can you share your better methods?

[Niels Henrik Bruun]

Fine solutions. All depends on what you want.

Using -collapse- might be what you want. Or -tab-

If you only have two categorical variables the command -sumat- in package matrixtools (ssc), but want the summary in a matrix

[fu gang]

So kind of you, Thank you very much , I'll try them soon