I was asked what do I do when the normal probability plot has a ‘S’ shape? one such as
What a good question, since this follows up on the methods to deal with non-normal data. I found four different conditions that could make that shape.
1. A 30000 ft metric that is increasing, such as revenue per month, or sales per month.
2. A 30000 ft metric that has the mean cycling between two values, such as temperature in a room.
3. A mixture of 30000 ft metrics that are in equal proportion and variance but with differen but close means, such as different operators or shifts.
4. A 30000 ft metric consisting of two equally occuring processes, such as day and night shift.
If you can think of any more, let me know. This investigation will be documented in the next technical report in the