# Technical Studies Reference

### Moving Average - Arnaud Legoux

This study calculates and displays an Arnaud Legoux Moving Average (ALMA) of the data specified by the Input Data Input.

Let $$X$$ be a random variable denoting the Input Data, and let $$X_j$$ be the value of the Input Data at Index $$j$$. Let the Inputs Length, Sigma, and Offset be denoted as $$n$$, $$\sigma$$, and $$k$$, respectively. Then we denote the Moving Average - Arnaud Legoux at Index $$t$$ for the given Inputs as $$ALMA_t(X, n,\sigma, k)$$, and we compute it for $$t \geq n - 1$$ as follows.

$$\displaystyle{ALMA_t(X, n, \sigma, k) = \frac{\sum_{j = 0}^{n - 1}\exp\left(-\frac{(j - \lfloor k(n - 1) \rfloor)^2}{2n^2/\sigma^2}\right) \cdot X_{t - n + 1 + j}}{\sum_{j = 0}^{n - 1}\exp\left(-\frac{(j - \lfloor k(n - 1) \rfloor))^2}{2n^2/\sigma^2}\right)}}$$

For an explanation of the Sigma ($$\Sigma$$) notation for summation, refer to our description here.

For an explanation of the Floor Function ($$\lfloor \cdot \rfloor$$), refer to our description here

ALMA is a weighted moving average with Gaussian weights. It is advertised as a Gaussian filter, however caution should be exercised in this interpretation. The Input $$\sigma$$ does not play the role of the standard deviation of the Gaussians. Rather, the standard deviation is determined by $$\frac{n}{\sigma}$$. The mean, or center, of each Gaussian is determined by $$\lfloor k(n - 1) \rfloor$$

#### Inputs

• Input Data
• Length
• Sigma: This Input controls the width of the Gaussian distribution of the weights.
• Offset: This Input controls the center of the Gaussian distribution of the weights.