Transformation types
Each transformation sequence is built from one or more of the following types. Understanding what each does helps you judge whether a suggested plan is appropriate.
Transformation reference
| Transformation | What it does | Example |
|---|---|---|
| Rename | Changes a column’s name without modifying its values. | RID → NACCID |
| Map | Replaces specific values in a column according to a defined mapping. | 1 / 2 → Male / Female |
| Add Suffix | Appends a tag to a column’s name when a variable only partially matches the target. | age → age_approx |
| Add New Category | Introduces a new permitted value to a standard column that does not yet include it. | Adding Unknown to a Yes / No column |
| Trim | Caps numeric values to a defined range; out-of-range values become the min or max. | Ages above 90 are set to 90 |
| Count | Creates a value by counting how many columns satisfy a condition. | Number of cognitive tests completed |
| Case When | Sets a column’s value from conditional logic on the same or other columns. | If score < 24 then Impaired, else Normal |
| Convert to Null | Replaces values that meet a condition with a missing value. | Placeholder -99 becomes blank |
| Formula | Computes a new value from a mathematical expression over existing columns. | BMI = weight / height² |
| Convert Data Type | Changes how values are stored without altering their content. | Text “5” → integer 5 |
| Convert Numerical to Categorical | Groups a continuous range into discrete labelled bins. | Age → <65 / 65–80 / >80 |
| Assign | Fills an entire column with a single fixed value for all rows. | Set source_study = AIBL for every row |
| Normalize | Cleans and standardizes text so values are consistent regardless of formatting. | “ Male”, “MALE”, “male” → male |
| Split | Divides text on a separator and retains the selected part. | “1990-05” split on –, keep year → 1990 |
| Group By Cumulative Count Unique | Orders rows within a group and numbers each sequentially, resetting per group. | Patient visits numbered 1, 2, 3… in date order |