Evaluating and Improving Models
Give every shift its own error report
Only one night reading remains
The night team sees errors that disappear inside the overall average. Reuse the residual calculation and keep shift membership attached.
BOTIE
The inputs we are comparing
Observed temperatures
Same in both cases
- 010
- 110
- 210
- 310
Predicted temperatures
Balanced shifts
- 09
- 19
- 29
- 39
A hidden night regression
- 09
- 19
- 29
- 36
Shift for each reading
Balanced shifts
- 0day
- 1night
- 2day
- 3night
A hidden night regression
- 0day
- 1day
- 2day
- 3night
Result for balanced shifts
[
{
"group": "day",
"n": 2,
"mae": 1
},
{
"group": "night",
"n": 2,
"mae": 1
}
]The night shift is now the majority
This check uses the pieces from “Preserve the night denominator”.
BOTIE
- Observed temperatures
- 04
- 18
- 26
- Predicted temperatures
- 03
- 15
- 26
- Shift for each reading
- 0night
- 1night
- 2day
Use a number, text, or an array matching the requested result. For a matrix, put each row in brackets. Saving your prediction runs your current construction on these inputs and checks the earlier examples too.
Build a construction in the workspace above first. This case will test those pieces.
Which new night observations would help distinguish a persistent problem from one unusual reading?
A slice needs both a clear definition and enough relevant evidence.
Counts and means alone do not quantify uncertainty, establish fairness or explain the cause of a group difference.
Work through the Python challenges →