Question about hat matrix X(X'X)^(-1)X'

  • Thread starter julie94
  • Start date
  • Tags
    Matrix
In summary: Yes, you can use the change of basis matrix P to simplify the expression. You can also use the fact that P^-1=P^T if P is an orthogonal matrix. From there, you can show that X(X'X)^-1X' = Y(Y'Y)^-1Y'.
  • #1
julie94
18
0
Hi everyone,

I am working on the following problem.

Suppose the set of vectors X1,..,Xk is a basis for linear space V1.
Suppose the set of vectors Y1,..,Yk is also a basis for linear space
V1.
Clearly the linear space spanned by the Xs equals the linear space
spanned by the Ys.

Construct an algebraic argument to show that
X(X'X)^(-1)X'=Y(Y'Y)^(-1)Y'

I am very confused, I am not sure what is meant by algebraic argument
in this instance, and I would welcome your ideas on how to tackle this
question.

Thanks in advance.
 
Physics news on Phys.org
  • #2
I am confused as to what is meant by "X" and "Y" here since you only mention, [itex]\{X_k\}[/itex] and [itex]\{Y_k\}[/itex] previously. But the reference to X', X-1, Y', and Y-1 indicate that X and Y are matrices or linear operators, not vectors. What is the relationship between X and [itex]\{X_k\}[/itex], between Y and [itex]\{Y_k\}[/itex]?

If X and Y are matrices or linear operators then:
It looks to me that once you use (AB)-1= B-1A-1 it becomes very easy to show that the two are the same!
 
  • #3
Sorry I meant to write

X=[X1: X2 :...: Xk]
Y=[Y1: Y2 :...: Yk]

You are right, X and Y are matrices. Thanks a lot for the help.
 
  • #4
I have tried (AB)-1= B-1A-1, but I am not getting what I need. Would you be kind enough to give me another hint.
 
  • #5
You want to show that [itex]X(X'X)^{-1}X'=Y(Y'Y)^{-1}Y'[/itex].

Actually, you can show much more than that.

As I said, "[itex](AB)^{-1}= B^{-1}A^{-1}[/itex] so that [itex](X'X)^{-1}= X^{-1}X'^{-1}[/itex] and then [itex]X(X'X)^{-1}X'= X(X^{-1}X'^{-1})X'[/itex]
Can you do that?
 
  • #6
I understand the lines you wrote. But I do not know where to take it from here.

What should I do with
LaTeX Code: X(X^{-1}Xsingle-quote^{-1})Xsingle-quote
?
Thanks a lot for the help.
 
Last edited:
  • #7
Do I need to write
X=PYP^{-1}
where P changes the basis from Y to X?
And plug in?
 

FAQ: Question about hat matrix X(X'X)^(-1)X'

1. What is the purpose of the hat matrix in statistics?

The hat matrix, also known as the projection matrix, is used to compute the predicted values of a linear regression model. It helps to identify the relationship between the independent and dependent variables by estimating the coefficients of the regression equation.

2. How is the hat matrix related to the correlation between variables?

The hat matrix is related to the correlation between variables because it is used to minimize the sum of squared errors in a linear regression model. It does this by projecting the response variable onto the space spanned by the independent variables, and the resulting vector of fitted values is then correlated with the original response variable.

3. What does the (X'X)^(-1) term in the hat matrix represent?

The (X'X)^(-1) term in the hat matrix is known as the inverse of the cross-product of the independent variables. It is used to compute the variance-covariance matrix of the estimated coefficients in a linear regression model.

4. How does the hat matrix help to detect influential observations?

The hat matrix helps to detect influential observations by identifying the observations that have a large impact on the estimated coefficients. These observations have high leverage, meaning they have a large influence on the fitted values of the regression model and can significantly change the results.

5. Can the hat matrix be used for non-linear regression models?

No, the hat matrix is specifically designed for linear regression models. It cannot be used for non-linear regression models as the assumptions and calculations are different. Other methods, such as the Jacobian matrix, are used for non-linear models to estimate the coefficients and detect influential observations.

Back
Top