I have the equation $RSS\left ( \beta \right )=\left ( y-X\beta \right )^{T}\left ( y-X\beta \right )$ that I'm trying to differentiate w.r.t. $\beta$. I used chain rule to get to $\frac{\partial RSS(\beta) }{\partial \beta } = (-X^{T})(y-X\beta)+(y-X\beta)^{T}(-X)$, but I don't know how to get from this to $X^{T}(y-X\beta)$. How do I deal with all the different transposes?


You can proceed as follows:

\begin{align} RSS\left ( \beta \right )&=\left ( y-X\beta \right )^{T}\left ( y-X\beta \right )\\ &=y^Ty - 2y^TX\beta + \beta^TX^TX\beta. \end{align}

Now, if you differentiate $RSS\left ( \beta \right )$ with respect to $\beta$, the derivative will be: $-2 X^T y + 2X^T X\beta $.

The results, which also can be found in any standard books, that I have used are:

  1. $\frac{d}{dx}x^TAx = 2Ax$, if $A$ is symmetric.

  2. $\frac{d}{dx}y^Tx = y$.

Here $x$ and $y$ denote two different vectors of the same dimension and $A$ is a square matrix of conformable dimension. Furthermore, $x^T$ indicates the transpose of the vector $x$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.