turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

03-19-2015 02:30 AM

Is there any function or algorithm to verity a matrix is nonsingular in IML? I need if in a "if then"statement.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to ShufeGuoding

03-19-2015 03:15 AM

The function DET returns the determinant of a matrix, so:

if det(x)^=0 then ......

but you may want to exclude small determinants if whatever you want to do is likely to give stability problems with near-singular x, so something like:

if abs(det(x))>s

where s is some suitable small value.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to ShufeGuoding

03-19-2015 08:23 AM

The problem you describe is equivalent to computing the "rank of a matrix." A nonsingular matrix is a square matrix that has full rank. Because of finite-precision arithmetic, the rank problem is difficult. As Ian points out, there are problems with trying to compare a floating-point computation (determinant, eigenvalue, singular value,...) with zero. Stable rank-detecting algorithms use some variation of Gaussian elimination to determine whether there are linear dependencies in the rows or columns of a matrix. In SAS/IML, the standard way to detect rank is to compute the generalized inverse (which always exists) and then multiply it by the original matrix. If you get the identity matrix, then the original matrix was full rank (nonsingular). Otherwise the original matrix is less than full rank. So to detect singularity, I suggest the following:

A = { /* your matrix to test */ };

matrixRank = round(trace(ginv(A)*A));

if matrixRank < nrow(A) then print "Singular";

else print "Nonsingular";