Yes indeed it is true that the slides explain the dropout reasons. The question is how does one handle the dropouts in the numerical analysis?
If one gives the dropout "0" scores for all future testing that they miss that artificially lowers the mean. But if one removes them completely than the mean may be artificially raised as it looks like 100% completed.
So one must decide how to represent the dropouts without skewing the data.
My post was meant to explain the methodology likely used to achieve this.
Apologies if that was not the point of your question.
Market Data 
Markets 
AI
