These are chat archives for FreeCodeCamp/DataScience
discussion on how we can use statistical methods to measure and improve the efficacy of http://freeCodeCamp.com
In my experience, it is a question of practicing. There are also cases where some approaches are more recommended than others. In that case it is better to follow what the majority of researchers do.
In case there are several approaches to tackle one thing, and let's say you are using supervised models, then there are techniques for model comparison.
Hope this helps.