Chapter 2 Statistics

2.1 Goals of statistical analysis

2.1.1 Description

2.1.2 Assessment of quality

2.1.3 Inference

2.2 General principles of statistical practice

2.2.1 Self-skepticism and humility

Try to find reasons to doubt your analysis, and be honest about them. Limitations matter.

2.2.2 Robustness

Do your analysis in several different ways and report the robustness (e.g., parametric and non-parametric). If the results depend strongly on how you approach the analysis, there may be a problem with the data.

Instead of opportunistic removal of outliers, try to choose robust analyses from the beginning.

2.2.3 Parsimony

Try to keep your analysis and visualizations simple and close to the data and question of interest.

2.2.4 Transparency

Remember that others understand much less about your data than you do. Try to present information in such a way that they, too, can understand the data.

• Don’t just present a test; present a visualization, or several.
• Avoid presenting complex transformations of the data if you can help it; presenting differences of differences, for instance, hides aspects of the data that may be important.
• Present the data, warts and all; do not remove apparent outliers from your visualizations.
• Clearly explain the choices you make in all phases of the analysis.

×