stl_diagnostics

stl_diagnostics(
    data,
    date_column,
    value_column,
    frequency='auto',
    trend='auto',
    robust=True,
)

Generate STL decomposition diagnostics (observed, season, trend, remainder, seasadj).

Parameters

Name	Type	Description	Default
data	pd.DataFrame or pd.core.groupby.generic.DataFrameGroupBy	Time series data, optionally grouped.	required
date_column	str	Name of the datetime column.	required
value_column	str	Numeric measure to decompose.	required
frequency	(str, int, float, pd.Timedelta, pd.DateOffset)	Seasonal period specification. `"auto"` (default) infers a period via :func:`pytimetk.get_seasonal_frequency`. Strings such as `"7D"` or `"30 days"` are supported.	`'auto'`
trend	(str, int, float, pd.Timedelta, pd.DateOffset)	Trend window specification. `"auto"` (default) infers a window via :func:`pytimetk.get_trend_frequency`.	`'auto'`
robust	bool	Apply a robust STL fit (down-weights outliers). Defaults to `True`.	`True`

Returns

Name	Type	Description
	pd.DataFrame	Decomposition with columns: - grouping columns (if present) - `date`, `observed`, `season`, `trend`, `remainder`, `seasadj`

Examples

import numpy as np
import pandas as pd
import pytimetk as tk

rng = pd.date_range("2020-01-01", periods=180, freq="D")
values = np.sin(np.linspace(0, 8 * np.pi, len(rng))) + np.random.default_rng(123).normal(scale=0.1, size=len(rng))
df = pd.DataFrame({"date": rng, "value": values})

decomposition = tk.stl_diagnostics(
    data=df,
    date_column="date",
    value_column="value",
    frequency="7D",
    trend="30 days",
)
decomposition.head()

	date	observed	season	trend	remainder	seasadj
0	2020-01-01	-0.098912	-0.306686	0.609350	-0.401577	0.207774
1	2020-01-02	0.103167	-0.223424	0.621052	-0.294462	0.326591
2	2020-01-03	0.405929	-0.003336	0.630854	-0.221589	0.409266
3	2020-01-04	0.428271	0.002081	0.638685	-0.212495	0.426190
4	2020-01-05	0.624586	0.065817	0.644524	-0.085755	0.558769