Construct Matrix P & Find Trace/Eigenvalues - Can Anyone Help?

Click For Summary

Homework Help Overview

The original poster attempts to construct a matrix P from a full rank nxp matrix X and subsequently find the trace and eigenvalues of P. The matrix is defined as P = I - X((X'X)^-1)X', where X' denotes the transpose of X. The context appears to relate to linear algebra and matrix theory, possibly in the setting of regression analysis.

Discussion Character

  • Exploratory, Conceptual clarification, Mathematical reasoning

Approaches and Questions Raised

  • Some participants discuss properties of matrix inverses and the implications for the construction of P. Others suggest that P is a projection matrix and raise questions about its dimensions and rank. Hints are provided regarding the relationship between the trace and eigenvalues.

Discussion Status

The discussion is ongoing, with participants exploring various properties of the matrix P and its implications. Hints and insights have been shared, particularly regarding the projection nature of P and the relationship between trace and eigenvalues, but no consensus or resolution has been reached.

Contextual Notes

There is mention of the matrix not needing to be square, which may affect simplifications. The problem appears to be situated within a broader context of multiple regression analysis, which may impose additional constraints or assumptions on the discussion.

abiadams
Messages
1
Reaction score
0
Am really lost on this one:

Suppose X is a full rank nxp matrix. Construct the matrix:

P = I - X((X'X)^-1)X'

Where X' is the transpose of X.

I then have to find the trace and eigenvalues of P.

Can anyone help?!
 
Physics news on Phys.org
do you know that for invertible matricies A,B that

[tex](AB)^{-1} = B^{-1} A^{-1}[/tex]
 
easy to see this, say C is the inverse of (AB), then

[tex](AB)C = ABC = I[/tex]

[tex]A^{-1}ABC = BC = A^{-1}[/tex]

[tex]B^{-1}BC = C = B^{-1}A^{-1}[/tex]
 
lanedance's comment is true, but since your matrix need not be square it doesn't help. You can't simplify

[tex] P = I - X(X'X)^{-1} X'[/tex]

any more than it already is.

This is the type of problem seen during a matrix introduction to multiple regression (although it may be in a different setting for your problem).

Hint: You can show that [tex]P^2 = P[/tex] ([tex]P[/tex] is actually a projection matrix).
1) What is its dimension?
2) If you know the rank of [tex]X[/tex] you can find the rank of [tex]P[/tex]

Those hints will (should, may) give you an idea of an attack to find the quantities you need.
 
ok, yeah - misread that sorry
 
along with statdads comments, the property of the trace being the sum of the eigenvalues
[tex]tr(A) = \sum_i a_ii = \sum_i \lambda_i[/tex]

and
[tex]tr(A^k) = \sum_i \lambda_i^k[/tex]
could help
 

Similar threads

  • · Replies 8 ·
Replies
8
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 24 ·
Replies
24
Views
4K
Replies
8
Views
3K
Replies
2
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
Replies
5
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 33 ·
2
Replies
33
Views
3K
  • · Replies 2 ·
Replies
2
Views
2K