Friends, I want to write my new posts about 2D Z-series in a programming context so you could use some main functions for powers and products of 2D Z-series and could also take their expectations in order to get jump-started with programming. But I will write detailed posts tomorrow.
Here are some notes copied from past but more detailed information tomorrow.
Two Dimensional Z-Series/Hermite-Series:
Case 1: Random variable X indepenent and Random Variable Y dependent and Conditional on X
Suppose we have two non-Gaussian continuous random variables, X and Y and we can make observations of their joint realizations. We suppose that X is the independent random variable and Y is the dependent random variable. This dependence might follow from causality but would usually be arbitrary for the analytics and we could possibly reverse the dependent and independent random variables and successfully re-do the analytics.
We can calculate the univariate Z-series/Hermite-series of independent random variable X as
\[\,X(Z_x)\,=\, ah_0 \, +\, ah_1 \,H_1(Z_x) \,+\, ah_2 \,H_2(Z_x) \,+\, ah_3 \,H_3(Z_x) \,+\,\ldots\]
Even though we can make a univariate Z-series/Hermite-Series of Y, this would not be helpful from two dimensional perspective. We just define a conditional version of Y as
\begin{align}
\left[Y|X\right](Z_y \, , \, Z_x)\, \,&=\, \Big[\, cH_{0,0} \, + \, cH_{0,1} \,H_1(Z_x) \,+\, cH_{0,2} \, H_2(Z_x) \,\, + \, cH_{0,3} \,H_3(Z_x)\,\,+\,\ldots\,\Big] \\
&+\, \Big[ \, cH_{1,0} \, + \, cH_{1,1} \,H_1(Z_x) \,+\, cH_{1,2} \, H_2(Z_x) \, + \, cH_{1,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_1(Z_y) \, \\
&+\, \Big[ \, cH_{2,0} \, + \, cH_{2,1} \,H_1(Z_x) \,+\, cH_{2,2} \, H_2(Z_x) \, \, + \, cH_{2,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_2(Z_y) \, \\
&+\,\,\,\ldots\, \\
\end{align}
Bivariate Joint Hermite-Series of X and Y is denoted by a random variable W in the form of a product (of joint realizations of X and Y) and is given as (please note that \(aH_{m,n}\) are merely numerical coefficients while \(H_n(.)\) are hermite polynomials)
\begin{align}
W(Z_y \, , \, Z_x)\, = \, X\, Y\,&=\, \Big[\, aH_{0,0} \, + \, aH_{0,1} \,H_1(Z_x) \,+\, aH_{0,2} \, H_2(Z_x) \,\, + \, aH_{0,3} \,H_3(Z_x)\,\,+\,\ldots\,\Big] \\
&+\, \Big[ \, aH_{1,0} \, + \, aH_{1,1} \,H_1(Z_x) \,+\, aH_{1,2} \, H_2(Z_x) \, + \, aH_{1,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_1(Z_y) \, \\
&+\, \Big[ \, aH_{2,0} \, + \, aH_{2,1} \,H_1(Z_x) \,+\, aH_{2,2} \, H_2(Z_x) \, \, + \, aH_{2,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_2(Z_y) \, \\
&+\,\,\,\ldots\, \\
\end{align}
The following version of conditional Hermite-Series would hold
\[ \,W(Z_y \, , \, Z_x)\, =\,\left[Y|X\right](Z_y \, , \, Z_x)\,\,X(Z_x)\,\]
Now, we want to do analysis of \(\left[Y|X\right](Z_y \, , \, Z_x)\, \). It is composed of three parts. The first part of Y is directly correlated with X and appears on first row in the formula of \(\left[Y|X\right](Z_y \, , \, Z_x)\, \). The second part of Y is that has variance independent of X and that appears in the first column in the formula of \(\left[Y|X\right](Z_y \, , \, Z_x)\, \). The third part of Y has variance that is dependent on X and this part appears below the first row and towards left of the first column. rewriting the formula of \(\left[Y|X\right](Z_y \, , \, Z_x)\, \), we have
\begin{align}
\left[Y|X\right](Z_x \, , \, Z_y)\, \,&=\, \Big[\, cH_{0,0} \, + \, cH_{0,1} \,H_1(Z_x) \,+\, cH_{0,2} \, H_2(Z_x) \,\, + \, cH_{0,3} \,H_3(Z_x)\,\,+\,\ldots\,\Big] \\
&+\, \Big[ \, cH_{1,0} \, + \, cH_{1,1} \,H_1(Z_x) \,+\, cH_{1,2} \, H_2(Z_x) \, + \, cH_{1,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_1(Z_y) \, \\
&+\, \Big[ \, cH_{2,0} \, + \, cH_{2,1} \,H_1(Z_x) \,+\, cH_{2,2} \, H_2(Z_x) \, \, + \, cH_{2,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_2(Z_y) \, \\
&+\,\,\,\ldots\, \\
\end{align}
The first row \( \Big[\, cH_{0,0} \, + \, cH_{0,1} \,H_1(Z_x) \,+\, cH_{0,2} \, H_2(Z_x) \,\, + \, cH_{0,3} \,H_3(Z_x)\,\,+\,\ldots\,\Big]\) has part of Y that is directly correlated with X.
And first column given as
\begin{align}
&+\, \Big[ \, cH_{1,0} \, \Big] \, H_1(Z_y) \, \\
&+\, \Big[ \, cH_{2,0} \, \Big] \, H_2(Z_y) \, \\
&+\, \Big[ \, cH_{3,0} \, \Big] \, H_3(Z_y) \, \\
&+\,\,\,\ldots\, \\
\end{align}
is independent of X while the remaining body given as
\begin{align}
&+\, \Big[ \, cH_{1,1} \,H_1(Z_x) \,+\, cH_{1,2} \, H_2(Z_x) \, + \, cH_{1,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_1(Z_y) \, \\
&+\, \Big[ \, cH_{2,1} \,H_1(Z_x) \,+\, cH_{2,2} \, H_2(Z_x) \, \, + \, cH_{2,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_2(Z_y) \, \\
&+\, \Big[ \, cH_{3,1} \,H_1(Z_x) \,+\, cH_{3,2} \, H_2(Z_x) \, \, + \, cH_{3,3} \,H_3(Z_x)\,\,+\,\ldots\, \Big] \, H_3(Z_y) \, \\
&+\,\,\,\ldots\, \\
\end{align}
is the part whose variance systematically depends upon X.
Friends, I had forgotten to take my tablets(kemadrin) that offset some of the side-effects of antipsychotic injections and therefore had a rather bad evening. I just took the tablets but I am feeling very tired. I want to explain things in the programming context so many friends start writing their own simple functions with 2D Z-series. I am going to sleep now but will write a new post tomorrow.