The Little Handbook of Statistical Practice (2012) (jerrydallal.com)

Article summary

The article discusses statistical practice, specifically the calculation of variance and the use of unbiased estimators. The resource explains that dividing by n-1 is the modern convention, but does not provide an intuitive explanation for this choice. The discussion revolves around the technical aspects of statistics and the importance of understanding the underlying mathematics. The article is part of a handbook that provides a concise guide to statistical practice.

Main themes

• statistical practice
• variance calculation
• unbiased estimators
• degrees of freedom
• statistical significance

What commenters say

• Dividing by n-1 is the unbiased estimator for variance, but dividing by n is the maximum likelihood estimator.
• The choice of estimator depends on the specific context and the trade-off between bias and mean squared error.
• The concept of using up a degree of freedom when calculating the mean is not intuitive and requires a deeper understanding of statistical mathematics.
• Asymptotically, the difference between estimators becomes negligible, but for small sample sizes, the choice of estimator can have a significant impact.
• The handbook's explanation of statistical concepts is clear and concise, but may not provide sufficient depth for all readers.
• The use of statistical significance and p-values can be misleading and requires careful interpretation.
• Neural networks with thousands of parameters pose unique challenges for statistical analysis and inference.
• The importance of understanding statistical concepts is not limited to data science, but is relevant to all fields that involve data analysis and interpretation.