Remove outliers

Remove outliers — `removeOutliers`

Drop or clip outliers in numeric columns.

Use cases

Strip data-entry spikes before averaging.
Winsorize (clip) extreme values to a sane range instead of deleting rows.

What it does

Computes per-column bounds (IQR, z-score, or percentile), then either drops rows that fall outside the bounds or clips their values to the boundary.

Before

name	age
Alice	28
Bob	250
Carol	35
Dave	-5
Eve	42

5 rows · 2 cols

Remove outliers (columns=[age], method=iqr, action=drop, factor=1.5)

After

name	age
Alice	28
Carol	35
Eve	42

3 rows · 2 cols

Configuration

Config key	Type	Required	Description
`columns`	string[]	Yes	Numeric columns to scan
`method`	string	No	`iqr` (default), `zscore`, or `percentile`
`action`	string	No	`drop` (default) or `clip` to the bounds
`factor`	float	No	IQR multiplier (default `1.5`)
`threshold`	float	No	z-score threshold (default `3.0`)
`lower` / `upper`	float	No	Percentile bounds (default `1.0` / `99.0`)

Each method has its own parameter: iqr uses factor, zscore uses threshold, percentile uses lower/upper (0–100).

Generated Python code

# method: iqr, action: drop
_s = df_1['amount']
_q1, _q3 = _s.quantile(0.25), _s.quantile(0.75)
_iqr = _q3 - _q1
_lo, _hi = _q1 - 1.5 * _iqr, _q3 + 1.5 * _iqr
df_2 = df_1[_s.between(_lo, _hi) | _s.isna()]

Tips & common mistakes

drop removes rows; clip keeps them and pulls outliers to the bound — choose based on whether row counts must stay stable.
Match the parameter to the method. Setting threshold while using iqr has no effect.
Inspect the effect with a histogram on the node's output.