Open Advanced Options with the gear in the model panel header. Each section's ? icon links here.

On (default): each unit's network is scaled to length 1, so units are compared on the pattern of their connections rather than on how many they made. Off: every network is scaled by the longest one, so units with more connections sit farther from the center. Applies to standard, transmodal, and ordered models.
The rotation decides which directions in the network space become the plot's axes.
SVD is used when there are no groups to rotate on. It finds the axes that explain the most variance between units.
Means Rotation places the X axis through the means of two groups, so the plot shows their difference as clearly as possible. It needs exactly two groups.
Rotation by Regression (generalized means rotation) finds the biggest differences between group means while controlling for the nesting of your unit columns. It works with two or more groups. Turning it off returns to the original ENA rotations.
What it controls for: each unit's network is regressed on the top unit
column (the groups) together with every intermediate unit column, with
their pairwise interactions, in a lasso regression that leaves the group
term unpenalized. The X axis is built from the group term alone, so
variation that belongs to the intermediate columns is kept out of it. With
units Condition > Session > Speaker, for example, the axis shows the
Condition difference after adjusting for Session. This is rENA's
ena.rotate.by.generalized() with the group column and the intermediate
columns as x_var. With only two unit columns there is nothing to control
for, and the rotation gives the same axis as a Means Rotation.
The adjustment changes the direction of the axis, not the number of units. The group statistics are still two-sample tests on every unit, so with nested units (the same speaker in several sessions) repeated observations of a person count as independent. Keep that in mind when reporting them.
Auto-update Rotations (on by default) sets these options for you from your unit selections, and re-sets them whenever the selections change. Turn it off to choose rotations yourself and keep them when you change units. Opening a set keeps the rotations it was saved with; they are only switched off if its groups no longer allow them.
Rotation Groups chooses the two groups the rotation is built on. They default to the first two groups in your units.
With Rotation by Regression on, Custom Regression lets you choose the variables yourself:
The lists offer the columns that have one value per unit: your unit columns and any other column that never changes within a unit, such as a pre/post score. A Numeric variable needs a number for every unit; if some units have none, exclude them or treat the variable as Categorical. A categorical X is compared between the two Rotation Groups; a numeric X uses every unit.
Turning Custom Regression off returns the regression to the top unit column and resets the Rotation Groups to its first two groups.

| Option | When it appears | What it does |
|---|---|---|
| Window Type | Standard models | Moving Window (default) or Infinite Window; Threaded isn't available yet. See Building a Model |
| Use end time | Time-based windows | Adds an End Time Column on the Windows tab |
| Modality | Transmodal models | Shows the modality column on the Windows tab |
| Show Ground / Show Response | Transmodal models | Show the ground and response columns |
| Show Connection Weights | Transmodal models | Adds a weight to each entry in the window tree |

See Transmodal Analysis.

Rotate this set with another set's rotation, so the two models share the same axes and their plots can be compared directly. The list shows the other sets in the project that use the same codes and the same network type (standard or ordered, with or without self-connections). Choose None to use this set's own rotation.
Chooses how each line's connections are counted before they are added up for a unit:
Weighted models aren't available for ordered networks.
Off by default. When on, each code's connection to itself (the code occurring
again within the window) becomes a connection of its own, written A & A. Six
codes then give 21 connections instead of 15. Self-connections are included in
normalization, rotation and node positions, and each one is drawn as a dot on
its code's node. The weighting is the same as R rENA's
accumulate(self_connections = TRUE).
Ordered networks always include self-connections, so there the switch shows on and can't be changed. Your choice is kept if you switch back to a standard network. A set with self-connections and one without can't share a rotation (see Use Rotation Matrix From). Simplify Model (PRIA) and the window-size recommendation still use the model without self-connections.
Opens the PRIA dialog; see Simplify Model.
Manage Derived Codes (or the forked-arrow icon on the Codes tab) opens the editor for compound codes (A & B) and nested subcodes (B ⊆ A); see Compound & Nested Codes.

Edit Code Masking Matrix lists every connection between codes. Click a check mark to switch that connection off (✕); the model is rebuilt without it. Standard models show one triangle of the matrix, since A–B and B–A are the same connection; with Self-Connections on, the diagonal (A & A) is included. Ordered models show the full matrix, with FROM and TO axes, so each direction can be switched separately. Reset to Default restores all connections; a Modified badge on the button shows that the mask has been changed.
Connections masked by a compound or nested code are shaded and can't be switched back on while the rule exists. Hover one to see which rule masks it, for example residual and subcode of Design Reasoning.
Significant Digits (0–6, default 2) sets how many decimals are shown in statistics.