Timeline for Is there a way to specify reduced-rank regression using $\mathbf{y} = \mathbf{X}\boldsymbol\beta + \boldsymbol\epsilon$?

Current License: CC BY-SA 3.0

15 events

when toggle format	what		by	license	comment
Feb 13, 2019 at 10:56	comment	added	pauljohn32		In Inzemann (1975) the matrices $X$ and $Y$ are turned sideways. GLM would have $N \times r$, one row per case, but the original RRR paper models Y as $r \times N$. Same with X. So rather than $X\beta$ in RRR $CX$. For me, that's confusing. It all works out, but notation is difficult comparing documents.
S Dec 18, 2017 at 13:34	history	bounty ended	Clarinetist
S Dec 18, 2017 at 13:34	history	notice removed	Clarinetist
Dec 12, 2017 at 14:32	history	tweeted			twitter.com/StackStats/status/940590289290452992
S Dec 12, 2017 at 14:24	history	bounty started	Clarinetist
S Dec 12, 2017 at 14:24	history	notice added	Clarinetist		Reward existing answer
Dec 12, 2017 at 14:05	history	edited	amoeba		edited tags
Dec 12, 2017 at 11:42	vote	accept	Clarinetist
Dec 12, 2017 at 10:10	comment	added	amoeba		@jbowman Yes, but it's not only this: see my answer.
Dec 12, 2017 at 9:06	answer	added	amoeba		timeline score: 4
Dec 12, 2017 at 7:32	history	edited	amoeba		edited tags
Oct 18, 2017 at 13:47	history	edited	Clarinetist	CC BY-SA 3.0	added 124 characters in body
Oct 18, 2017 at 13:44	comment	added	jbowman		It's more likely because Izenman's definition of $x$ does not include a constant term, so he needs to subtract the mean from $y$ in order to have the mean of $\epsilon = 0$. The way to change it into the first form is to add a column of $1$s to $x$ and expand the coefficient matrix to include one more term..
Oct 18, 2017 at 13:29	history	edited	Clarinetist	CC BY-SA 3.0	added 7 characters in body
Oct 18, 2017 at 13:24	history	asked	Clarinetist	CC BY-SA 3.0