Please help me formulate a multivariate quantile function to facilitate the random sampling of a distribution. I have an idea, and I want to get feedback if it makes mathematical sense.

Let there be the continuous random variables $X$ and $Y$.

Let us assume we are given the PDFs $f_{X,Y}$, $f_X$, and $f_Y$.

We can then define the conditional CDFs... $$F_{X|Y}(x, y) = \mathbb P(X \le x | Y = y) = {1 \over f_Y(y)} \int_{- \infty}^x f_{X, Y}(t, y) dt$$

$$F_{Y|X}(y, x) = \mathbb P(Y \le y | X = x) = {1 \over f_X(x)} \int_{- \infty}^y f_{X, Y}(x, t) dt$$

(See https://mathcs.clarku.edu/~djoyce/ma217/conddist.pdf for explanation of above formulas)

We can then define the quantile function $\omega_{X, Y}$ in the following way... $$\omega_{X, Y}(p_1, p_2) = (x, y)$$ $$\text{where }p_1, p_2 \in [0, 1]$$ $$\text{such that...}$$ $$F_{X|Y}(x, y) = p_1$$ $$\text{ and }$$ $$F_{Y|X}(y, x) = p_2$$

Is this a mathematically valid way to construct a multivariate quantile function?


That will not work (unless $X$ and $Y$ are independent) and indeed it may not give you a way into the calculation as you probably need to know $y$ to solve $F_{X\mid Y}(x \mid y) = p_1$ for $x$, but you also need to know $x$ to solve $F_{Y\mid X}(y\mid x) = p_2$ for $y$

But it is not far away. Instead, try solving

$$F_{X}(x) = p_1$$ for $x$, and then use that $x$ to solve $$F_{Y\mid X}(y \mid x) = p_2$$ for $y$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.