The Student Room Group

Matrices (Generalized inverse)

By using matrix AA, derive the result x=COV(x,y)VAR(x)x = \dfrac{COV(x,y)}{VAR(x)} (Covariance and Variance) , c=yˉmxˉc = \bar{y} - m\bar{x} using generalised inverse methods, i.e.

m=(ATA)1Atγm =(A^T A)^-1 A^t \gamma where m=(mc)m = \begin{pmatrix} m \\ c \end{pmatrix}
Any ideas on how I'd go about doing this?
Original post by Jagwar Ma
By using matrix AA, derive the result x=COV(x,y)VAR(x)x = \dfrac{COV(x,y)}{VAR(x)} (Covariance and Variance) , c=yˉmxˉc = \bar{y} - m\bar{x} using generalised inverse methods, i.e.

m=(ATA)1Atγm =(A^T A)^-1 A^t \gamma where m=(mc)m = \begin{pmatrix} m \\ c \end{pmatrix}
Any ideas on how I'd go about doing this?


A little bit of context would help here! I presume that you're trying to derive the least squares estimator in linear regression and that you really mean

m=COV(x,y)VAR(x) \displaystyle m = \dfrac{COV(x,y)}{VAR(x)}

Is that right?

If so, this is pretty standard bookwork, and you can find some hints in places like this.
Reply 2
Original post by Gregorius
A little bit of context would help here! I presume that you're trying to derive the least squares estimator in linear regression and that you really mean

m=COV(x,y)VAR(x) \displaystyle m = \dfrac{COV(x,y)}{VAR(x)}

Is that right?

If so, this is pretty standard bookwork, and you can find some hints in places like this.


Sorry, the actual question is:
"We have showed on the least-squares worksheet that the best fit straight line through the data is given by

m=COV(x,y)VAR(x),c=yˉmxˉ \displaystyle m = \dfrac{COV(x,y)}{VAR(x)}, c = \bar{y} - m\bar{x}

By using the matrix A, derive this result using generalised inverse methods.

A=(x1x2x3111)A = \begin{pmatrix} x1 & x2 & x3 \\1 & 1 & 1 \end{pmatrix}
Reply 3
Any ideas anybody?
Original post by Jagwar Ma
Any ideas anybody?


To give you an answer suitable for your needs, it would really help to have some context: what have you been taught about generalized inverses, for example? Which course does this question come from?

The theory of how the generalized inverse minimizes squared error can be found here, for example -- or here, with a little more focused context for the problem in hand.
(edited 7 years ago)
Reply 5
Original post by Gregorius
To give you an answer suitable for your needs, it would really help to have some context: what have you been taught about generalized inverses, for example? Which course does this question come from?

The theory of how the generalized inverse minimizes squared error can be found here, for example -- or here, with a little more focused context for the problem in hand.


I'm doing a data modelling course and have been set a problem sheet. This is just one of the questions from the sheet. I can't really provide more context then that..All I can say is that the questions I'm set are often rather obtusely written and aren't stuff you can simply look up..it's a common theme my lecturer likes to abide by so that problem sheets are particularly challenging.

Quick Reply

Latest