Demonstrating a form of E(SSR) via quadratic forms

I've been given a problem asking that I show that, in a normal error regression model, $$E(SSR)=σ^2+β'X'(I_n-(1/n) J_n)Xβ$$ I am new to applying matrix algebra to statistics, but I do know that the quadratic form of SSR is $$SSR=Y'[H-(1/n)J]Y$$ And after a few videos (and reading cross-validated) I understand why $$E(Y'AY)=tr(AΣ)+(μ'Aμ).$$ My attempted work so far (please correct me if I'm wrong) is that $$μ=E(Y)=Xβ$$, so my understanding is that I can rewrite $$μ'=Y'=β'X'$$ and $$μ=Y=Xβ$$, which gets me to $$E(Y'AY)=trace([H-(1/n)J]Σ)+β'X'[H-(1/n)J]Xβ$$ Just based on the pattern, I could hazard a guess that $$trace([H-(1/n)J]Σ)=σ^2$$, but I can't figure out why. Moreover, I don't see how I can make $$β'X'[H-(1/n)J]Xβ$$ become $$β'X'(I_n-(1/n) J_n)Xβ$$. Perhaps all of this is trivial, but I'm new to it and it was not really explained prior to the question being asked.

EDIT: I've presented this exactly as it was presented to me. I believe $$J_n$$ refers to an $$n$$ x $$n$$ matrix filled with 1's. I would assume that $$n$$ is the same for all $$J$$ and $$H$$ mentioned, but there was no additional detail provided.

• You should define $J$, $J_n$, $H$ – Christoph Hanck Oct 12 at 10:12
• I presented the required solution as it was given to me. I believe $J$ and $J_N$ refer to a matrix of 1's and an $n$x$n$ matrix of 1's, respectively. I would assume that $n$ is the same for the entire problem, but it was not specified. – AJV Oct 12 at 12:20
• I find it difficult to start here. $H$ might, in common notation, be the hat matrix $X(X'X)^{-1}X'$, but then a common way to write $SSR$ is $Y'(I_H)Y$, which makes me wonder how the above result comes about. So we will need more context, I believe – Christoph Hanck Oct 13 at 5:02