Assuming that the universe is knowable to some degree, a useful prediction involves dynamic modelling, data analysis and human judgement. How specific the modelling, how accurate any resulting data might be, and whether predictions can be constructed as a series of conditions informing the next are all important to know from a statistical perspective. There are also non-statistical questions worth asking before investing significant resources in predictions, such as: What is at stake? What might be the consequences of different outcomes? Is there a way to establish conditions where any outcome will be of benefit? Which things are not being modelled? Even once these questions have been answered, it can still be misleading to assume predictions are facts on which decisions can be made. Yet we are creatures of habit, making small or significant decisions which collectively form patterns. These patterns produce data which can be analysed to better understand the patterns. The ability to separate things that can be easily (or usefully) predicted to provide 'actionable intelligence' from those that are too messy, complex or unknowable lies at the heart of any attempt at modelling or analysis. This should be kept in mind while taking any data-driven approach.

• Model making and asking the right questions
• Regression analysis
• Statistical inference
• Nonparametric methods
• analysis and judgement
• …