MAD rescaled by 1.4826 is comparable to σ under a Gaussian distribution

It comes from the equation: $\sigma /MAD\approx 1/0.6745\approx 1.4826$

Since 𝛔 is used in many functions, the rescaled MAD can serve as a drop-in replacement for 𝛔, but unlike it, it won’t be affected by the outliers. This applies only if data follows a Gaussian distribution.