Skip to main content

15  Configuration

Every upload, including a Data Extraction table, uses the same column editor. It opens with guessed column roles and groups unless a saved profile matches the headers exactly. Check the guesses, then use them for this upload or save them as a reusable profile.

This walkthrough uses iris.csv, a table of 150 flowers and six columns.

1. Upload Iris

Upload iris.csv on Data Analysis. No saved profile matches it yet, so the column editor opens with each column’s role and group guessed from its name and values. Plots appear after you choose Use for this upload or save a profile. The guide at the top, How to customize Data Analysis for your data, summarizes the editor; click its title to fold it.

The column editor for iris.csv: an expanded guide, the Bulk Column Role Assignment and Feature Group Management controls, a table of the six columns with guessed roles and groups, and the Use for this upload and Save profile as buttons.

The column editor as it opens for iris.csv.

For your own data, use column names on the first row and one observation per row. The uploader accepts CSV, TSV/TXT with tab, semicolon, or pipe separators, Excel .xlsx/.xlsm, and OpenDocument .ods. Put a spreadsheet’s table on its first sheet.

2. Review column roles

The editor reports 6 columns × 150 rows. Each column has one Role: Row ID, Categorical, Numerical, or Ignore. See Column roles for their uses and requirements.

For Iris, keep the four length and width columns Numerical, set species to Categorical, and assign flower_id as Row ID:

Role dropdowns assign sepal_length, sepal_width, petal_length, and petal_width to Numerical, species to Categorical, and flower_id to Row ID.

The six Iris columns with flower_id assigned as Row ID.

Use each column’s Role dropdown to change its use. For example, a numeric treatment code can be Categorical so it appears in filters. Changing a column away from Numerical removes its feature-group assignment.

To change several columns at once, use Bulk Column Role Assignment: tick their rows, choose a role beside Ticked columns are, and click the button that names the change, such as Make 2 Categorical or Ignore 2. Select all ticks every row except the Row ID, which shows a lock instead of a tick box; change it with its own Role dropdown.

Check flagged guesses

When a guess looks doubtful, ⚠️ Check these guesses appears above the bulk controls. Each line explains the doubt and offers a one-click fix, and the column’s preview is marked ⚠️. Iris has none. In the sample wine_quality.csv, quality holds only six whole numbers, so it may be a group label rather than a measurement:

Check these guesses (1): quality (Numerical): 6 whole-number values in 1,599 rows. If it labels groups such as day, well or plate, make it Categorical. A Make Categorical button sits at the right.

A whole-number score flagged as a possible category.

Other flags catch text columns with too many distinct values for a useful filter, and numbers stored as text, such as decimal commas, units, or placeholders like below LOD. Their fix is Ignore, or correct the file and upload it again. Applying a fix clears its flag; columns from a saved profile are never flagged.

A Row ID is optional. Automatic suggestions use exact identifier names from extraction profiles, such as cell_id; you can assign another column manually. Only one column can be Row ID, and its values must be nonmissing and unique. Assigning a new ID returns the previous ID column to Numerical or Categorical according to its values.

With no assigned ID, the app adds Row number, populated with 1 through the number of input rows. If that header already exists, it uses Row number.1, then .2, and so on. These labels follow input row order, so use your own identifier when identity must remain stable across reordered files. For Iris, change flower_id to Ignore to use generated row numbers.

At least one Numerical column must contain usable numbers. Field-of-view columns such as image_name use the Categorical role in every upload. Categories are stored as text; whole-number labels use 1 rather than 1.0, and missing values become N/A.

3. Feature group management

Feature groups split Numerical columns into shorter picker lists without changing their values. Unassigned measurements appear under Uncategorized Features. Feature groups organize measurements; categorical columns group or filter observations.

How group names are inferred

Recognized extraction headers form groups by extractor and channel, such as Lifetime fit_nadh. Columns beginning with Derived: form Derived Features. These rules apply to any uploaded table.

For other headers, at least two new Numerical columns must share the prefix before the earliest _, ., or : (colon and space) to form a new group:

Column headers Suggested group
sepal_length, sepal_width sepal
petal_length, petal_width petal
NADH.mean, NADH.std NADH
Texture: contrast, Texture: entropy Texture

Hyphens stay within names: anti-PD1_mean and anti-PD1_std form anti-PD1. Saved assignments are preserved, including renamed groups and ungrouped columns. New columns can follow matching saved columns into their group or join an existing group with the matching name.

Change groups in bulk

Keep the suggested sepal and petal groups, or combine the four measurements into Flower size:

  1. Tick the four measurement rows.
  2. Click ➕ New group, type Flower size, and click Create with 4.

The four Iris measurements are ticked; their groups are still sepal or petal. The New group box holds the name Flower size and a Create with 4 button. flower_id, the Row ID, shows a lock instead of a tick box.

Four ticked measurements about to form the Flower size group.

To move ticked measurements into an existing group, choose it beside Ticked Numerical go in group and click Move 4. Choose Uncategorized and click Ungroup 4 to remove assignments. Ticked rows that are not Numerical are skipped. For one feature, change its row’s Group directly.

To rename or delete a group, choose it in the same dropdown and click ✏️. Rename carries its columns over; using an existing name merges the groups. 🗑️ Delete group leaves its measurements uncategorized.

Flower size is chosen beside Ticked Numerical go in group. The open pencil box shows Rename to Flower size, a Rename button, and a Delete group button.

The ✏️ box acts on the group chosen beside it.

4. Use or save your choices

  • Use for this upload applies the reviewed roles and groups without changing saved profiles.
  • 💾 Save profile as stores a new profile and resumes analysis. When starting from guessed roles, enter Iris beside this button to reuse these choices later.
  • 💾 Save to {profile} & use appears when a saved profile is applied. It updates that profile and resumes analysis.

If the name entered beside Save profile as already exists, confirm with ⚠️ Replace profile or enter another name. Column edits take effect only when you choose Use or Save.

5. Reopen the column editor

Select sepal_length from Flower size to begin analysis, or from sepal if you kept the automatic groups.

If you saved the example as Iris, the summary reads Iris: 1 Categorical · 4 Numerical; counts omit the optional Row ID. Click ✏️ Fix Categorical & Numerical features below it to reopen the editor.

Iris: 1 Categorical · 4 Numerical appears above a button labeled Fix Categorical & Numerical features. The Flower size picker has sepal_length selected.

Reopen the column editor from the button below the summary.

Temporary edits show This upload only, or {profile} · edited for this upload when based on a saved profile. Reopening hides the analysis controls and plots while preserving their settings.

Choose Use for this upload or Save to apply further edits. Cancel restores the choices in use before reopening, including earlier upload-only edits, and resumes analysis.

6. Reuse and manage saved profiles

Automatic profile matching requires exactly one match to the complete header set, including columns marked Ignore. Case, spaces, and punctuation matter; column order does not. Otherwise, the editor opens with guessed roles and groups. A matched profile still opens the editor if a column needs attention, such as an assigned ID with repeated values.

Inside the editor, Apply a saved profile (optional) lists related profiles with shared, missing, and new column counts. Applying one replaces the current draft and keeps the editor open. It preserves saved roles and groups for present columns and guesses new ones. Saving back to that profile records new columns and removes absent ones.

Expand Manage saved profiles below the action buttons to rename (✏️) or delete (🗑️) any saved profile. These changes take effect immediately; Cancel only discards column and group edits. Deleting the profile supplying the draft resets it to the guessed roles.

You can store 20 profiles. At the limit, replace an existing profile or delete one before saving another.