Link between variance and Discrete Fourier Transform

522 Views Asked by At

I've read that:

$$ var(r) = \frac{1}{T}\sum_{t=0}^{T-1}(r_{t}-\bar{r})^2$$

is equal to:

$$ var(r) = \frac{1}{T^2} \sum_{k=1}^{T-1} R_{k}R_{k}^*$$

with $R_k$ the $T$-point DFT coefficients of $r_t$:

$$ R_{k}= \sum_{t=0}^{T-1}{r_t}e^{-j\frac{2\pi kt}{T}}$$

I tried to play with the formulas to match the equations, but without success. Any hint on the solution?

1

There are 1 best solutions below

0
On

(let us assume that we work on a centered function (mean = $0$).

Nothing astonishing : the continuous Fourier Transform is an isometry :

$$\|{\frak FT}(r)\|=\|r\|$$

This property is known to be preserved for the (DFT) Discrete Fourier Transform (see Parseval/Plancherel theorems there; morever, if you observe the variance, it is defined as a sum of squares, thus is by itself a norm (in the hermitian sense) ; from there your formula follows.

(don't pay too much attention to coefficient $1/N$ ; it is possible to eliminate it into by tuning the definition of the DFT).