Since many questions can either resolve Yes-early or No-late, it will look like you have a Yes bias until you reach the time the long questions resolve No.
The solution is to only include questions in calibration once they've reached their scheduled resolve date.
There should be a note to that effect below the calibration curve.
The questions can still be included in the other graphs for now.
Since many questions can either resolve Yes-early or No-late, it will look like you have a Yes bias until you reach the time the long questions resolve No. The solution is to only include questions in calibration once they've reached their scheduled resolve date. There should be a note to that effect below the calibration curve. The questions can still be included in the other graphs for now.