Thank you for the most interesting and high quality post I have read on Steemit. A good little snippet of data or research makes for a good post on platform such as this.

Attrition, which is sometimes called survivor bias, is perhaps one of toughest factors to consider when working with data such as this. Knowing this is also information and can lead to discoveries by examining why there is attrition. Statistics isn't enough when building a model but an understanding of the processes that drives the output and changes should pursued. Often referred to as the DGP, data generating process, should be known but most often isn't fully. Even some data from the FED is unusable if you don't know the origin of the data and policy changes overtime. Many data sets are poorly documented and the authors unwilling to engage in a discussion about it.

My favorite recent interesting anecdotes from research along these lines is that they had writings for nuns when they first entered the church and also know their mental outcomes in old age. Researchers say an analysis of writing of nuns who develop Alzheimers show less complexity in their than others at an early age. Suggesting the onset begins much earlier than any symptoms can be identified.

Great Post++++++

Navaman