I have a hang on how pi does time weighted averages.
However, it is not clear how this works in situations involving
- Interspersed bad values. I assume that first, the window is validated based on pctgood constraint. If it passes that constraint, the bad values are skipped and the time weighted average would be calculated on good values as if the bad values did not exist. Is this correct ?
- Boundry values. How far back or forward in time relative to the timewindow does pi look to find a boundry value.
- What if the nearest boundry value is a bad value ? how does this contribute to pctgood constraint ?