1
$\begingroup$

I need some results on expected roots of polynomial with random coefficients. I searched for $\textit{roots of polynomial with random coefficients}$ online and found some papers, but they are dealing with more complicated problems than I need. I think my particular problem is more specific: suppose we have a, say, cubic polynomial equation $$p_1(x) = ax^3 + bx^2 + cx + d = 0$$ where $a,b,c,d$ are some deterministic real numbers.

Suppose that a polynomial with random coefficients (in my case, the highest degree coefficient is not random, not sure if this will make a difference or not) is $$p_2(x) = ax^3 + \tilde{b}x^2 + \tilde{c}x + \tilde{d} = 0$$ where $\tilde{b}, \tilde{c}, \tilde{d}$ are random such that $\mathbb{E}[\tilde{b}] = b, \mathbb{E}[\tilde{c}] =c, \mathbb{E}[\tilde{d}] = d$, and $\tilde{b}, \tilde{c}, \tilde{d}$ have known and finite variances. I wonder, if $x$ is a root to $p_1$, and $\tilde{x}$ is a root to $p_2$, then can we say anything on the relation between $x$ and $\mathbb{E}[\tilde{x}]$? Is $\mathbb{E}[\tilde{x}] = x$? If not, can we write $\mathbb{E}[\tilde{x}]$ in terms of the variance of coefficients?

Since there are multiple roots of a cubic polynomial, it may make above statements sound strange. But the basic idea is, if the polynomial with deterministic coefficients has a root at $x$, do we expect the corresponding polynomial with randomness in coefficients also have a root at $x$? Many thanks!

$\endgroup$

1 Answer 1

2
$\begingroup$

Is $\mathbb{E}[\tilde{x}] = x$?

No --- in the case of just $x^3-d$, the root is $\sqrt[3]{d}$, and $E(\sqrt[3]{d})\ne\sqrt[3]{E(d)}$ typically. For instance let $d$ be 1, 8, 27 with probability $1/3$ each, then $$E(\sqrt[3]{d})=2\ne\sqrt[3]{12}=\sqrt[3]{E(d)}.$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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