Simple linear regression model and its column rank

227 Views Asked by At

Consider the simple linear regression model

$$Y_i = \beta_0 + \beta_1 X_i + \epsilon_i, i = 1, \dots, n$$

Give the order and column rank of the following for the model

$Y:$ Order = ? Column Rank = ?

$X:$ Order = ? Column Rank = ?

$\beta$ Order = ? Column Rank = ?

$X' X$ Order = ? Column rank = ?


Attempt

I believe $Y$ would be a n x 1 matrix

$X$ is going to be a n x 2 matrix

I don't know what $\beta$ means

$X'$ is 2 x n matrix

so $X'X$ would be a 2 x 2 matrix

therefore

$Y:$ Order = $n \times 1$ Column Rank = $1$

$X:$ Order = $n \times 2$ Column Rank = $2$

$\beta$ Order = ? Column Rank = ?

$X' X$ Order = $2 \times 2$ Column rank = $2$

Am I right? And how do I get $\beta's$?

1

There are 1 best solutions below

0
On BEST ANSWER

I assume you are referring to the rows and columns of the matrices/vectors which is referred to as the format of a matrix.

As far as I can see you are trying to determine the linear regression for a data set $X$ and a weight vector $\beta$ normally. We use the data set $\tilde{X}$ and add an additional column to the left which only contains one entry. This manipulation is done to get a vector equation.

$$\mathbf{y}=\mathbf{X\beta}+\mathbf{\varepsilon}$$

Please note that $\beta$ and $\varepsilon$ are vectors it is a strange behavior of LaTeX/MathJax. The output vector $y$ has the format $n\times 1$ containing the outputs/predictions for each observation. If we have $p$ predictors then $\mathbf{X}$ has the format $n \times (p+1)$ (note that we have an additional column for the added column) and $\beta$ has the format $(p+1)\times 1$. The error has the format $n\times 1$.

Can you determine the format of $\mathbf{X}^T\mathbf{X}$ as $(p+1) \times (p+1)$.