Add a square*:-
For the housing data find a best fit of the form Price ≈ a + b Age + c Sqft + d Sqft2. (Linear estimation entropy*) Suppose that an unknown parameter x has a normal (Gaussian) distribution with variance σ2. Independent measurements of the form yi = x +εi are made, where each εi is Gaussian with mean zero and variance ω2 and is independent of x.
The entropy of a multidimensional Gaussian variable is H = ½ log (2πe) n|Q| bits, where |Q| is the determinant of the covariance matrix corresponding to the variable. (See chapter 22 for the n = 1 case.)
(a) Find I(y1, y2, ... , yn; x)
(b) Let
And find I( ; x), showing that for purposes of estimation it is only necessary to consider y rather than keeping track of the individual yi's.