I do not teach standard deviation (I teach mean deviation instead)

Xavier Bordoy


"currículum" "desviació mitjana" "desviació típica" "estadística" "bitàcola"


Why we have to teach standard deviation?

Legally, I have to teach standard deviation σ\sigma. But really I have to teach standard deviation? In low levels of secondary education I do not teach it. I teach mean deviation, dmd_m. Why:

  • It’s more intuitive: if we want to mesure the regularity, the homogeneity of a sample, then I think that it’s more intituitive to calculate the mean of the diferences between mean and the values of the sample instead of calculate the square root of those differences squared. All my students say “Why squared?”
  • We don’t need standard deviation until high courses. For examples if we teach gaussian distribution, confidence intervals or linear regression. But if we introduce statistics or if we want to analyze data samples, I think we need simple deviation parameter1

The only disadvantage I think we would have if we teached the dmd_m instead of σ\sigma is that we could not take the most references in scientific literature, which use σ\sigma. I thought about it many times. And I think that the reason, the real reason, for which mathematicians use σ\sigma instead dmd_m is because, as a function, σ\sigma it’s derivable and dmd_m is not. So it allows getting easy bounds with known distributions (like gaussian ones).

  1. A range of the sample could be another one↩︎