Saving nonzero matrix elements

[Bill Smith]

I have a matrix that looks like this:

[0 x x x 0 x x x x]


where the x's are nonzero elements. I'd like to retain those nonzero elements in another matrix. I looked at a couple of things that were either close http://www.statalist.org/forums/foru...-of-the-matrix or dated http://www.stata.com/statalist/archi.../msg00136.html. Can anyone provide me guidance on the best way to accomplish this task?

[Nick Cox]

Do you mean matrix or is it sufficient to tell you about row vectors?

[Bill Smith]

I want to save that matrix into another matrix like so:

[x x x x x x x ]

Leave out the zeros.

[Nick Cox]

You haven't answered my question, but

 
. mata

: x = (0,1,0,0,2,0,0,0,3)

: select(x, (x :> 0))
       1   2   3
    +-------------+
  1 |  1   2   3  |
    +-------------+

[Bill Smith]

Well, I thought it was obvious it was a row vector, but thanks nonetheless. Using Mata introduces some other problems. I need to know how to pass the original vector into Mata and the reduced vector out. I've tried a couple of things I've seen online without success. Still looking at this point.

[Nick Cox]

It was clear to me, and no doubt many others, that your example was a row vector but your question was about a matrix. It's not so obvious what's the result of omitting zeros from a matrix is to be, for example, do you omit rows, columns, or both that contain 0?

Otherwise, looking in the help for Mata -- index and guide to functions -- Stata interface functions guides you to st_matrix().

 
. mat x = (0,1,0,0,2,0,0,0,3)

. mata : st_matrix("x", select(st_matrix("x"), st_matrix("x") :> 0))

. mat li x

x[1,3]
    c1  c2  c3
r1   1   2   3

[Bill Smith]

Ok. Was using st_matrix. And, yes, I looked through the Mata help. I'm just not a Mata programmer, so the learning curve is steep. Thought I needed an end statement, but that was causing my program to terminate. I'll give this a try. Thanks again.

[Bill Smith]

Thought I could get away with working with vectors, but found out today I'm going to need to perform this operation on a covariance matrix. It will be lower triangular and have combinations of rows and columns of zeros I want to delete. For example, I could have:

 
0










0
1









0
1
1








0
1
1
1







0
1
1
1
1






0
0
0
0
0
0





0
1
1
1
1
0
1




0
1
1
1
1
0
1
1



0
1
1
1
1
0
1
1
1


0
1
1
1
1
0
1
1
1
1

In this case I'd like to retain:

 
1








1
1







1
1
1






1
1
1
1





1
1
1
1
1




1
1
1
1
1
1



1
1
1
1
1
1
1


1
1
1
1
1
1
1
1

How would I need to modify the code to perform this operation? Thanks in advance.

[Bill Smith]

Someone explain this. I have matrix v with lower triangular form like above, except real number rather than ones. I can list its contents by saying:

matrix list v

Now, if I say:

mata : submat1=select(v, (0\1\1\1\1\0\1\1\1\1))

I get:

<ismt>:  3499  v not found

r(3499)

and termination. So it's not finding the matrix input. Yet it's clearly there. I'd appreciate some insight as to what the problem is and how I might solve it.

[Nick Cox]

Mata doesn't know about Stata's matrices, or indeed vice versa. That's the reason for st_matrix(), which you were using earlier.

[Bill Smith]

I tried:

mata : st_select(st_matrix("v1"), st_matrix("v"), (0\1\1\1\1\0\1\1\1\1))

but that produced a conformability error. v is a 10X10 matrix, so I don't understand.

[Nick Cox]

I'd move this over to the Mata forum and start a new thread.

One of the main reasons for the separate existence of the Mata forum was so that StataCorp developers could spot Mata questions easily.

[Bill Smith]

Good idea. Thanks for your feedback.