Largest complete panel obtainable under a coverage requirement.
A blocked design needs a complete matrix, but entries that joined a campaign late are scored on fewer periods than the rest. Admitting them costs periods for everyone, and excluding them costs entries. This helper makes that trade explicit and repeatable: keep the entries scored on at least min_periods periods, then keep the periods scored for all of them.
Raising min_periods admits fewer entries over a longer window, lowering it admits more entries over a shorter one. Neither direction dominates, so a campaign report is usually clearer with two or three panels bracketing the trade than with one.
This is pure: no logging, no plotting, no mutation.
| Name | Type | Description | Default |
|---|---|---|---|
| matrix | pd.DataFrame | Wide period-by-entry frame, one row per period and one column per entry, with NaN where an entry was not scored. | required |
| min_periods | int | None | Minimum number of scored periods an entry needs to enter the panel. Defaults to None, which admits every entry and therefore keeps only the periods scored for all of them. | None |
| Name | Type | Description |
|---|---|---|
| pd.DataFrame | pd.DataFrame: Complete sub-frame with no NaN, carrying the | |
| pd.DataFrame | admitted entries as columns and the surviving periods as rows. |
| Name | Type | Description |
|---|---|---|
| ValueError | When min_periods is negative, or the requirement admits no entry or leaves no period. |
import numpy as np
import pandas as pd
from spotforecast2_safe.stats.comparison import complete_panel
days = pd.date_range("2026-06-10", periods=6, freq="D")
matrix = pd.DataFrame(
{
"early": [500.0, 510, 505, 495, 515, 500],
"late": [np.nan, np.nan, 520.0, 512, 530, 515],
},
index=days,
)
print(complete_panel(matrix).shape) # 2 entries, 4 days
print(complete_panel(matrix, min_periods=6).shape) # 1 entry, 6 days(4, 2)
(6, 1)