Analyze

The Quantitative panels turn a selection into numbers: distributions and scatter plots, correlation and covariance matrices, point-to-point similarity, group statistics, and a Python notebook wired into the workspace. Every panel in this group is opened from View → Quantitative.

How these panels get their data

Almost all of them read a bag, not a raw selection, and almost all of them read only the most recently selected bag. If a panel looks empty, the usual cause is that no bag is selected, or that the bag you want is not the last one you clicked. Two panels are exceptions: Meaningful Features reads the live selection, and Statistics describes the whole space rather than any selection.

Read nextBagsHow to create and select the bags these panels consume.

Opening one for the current selection

Four of these panels register a point action, so you can reach them by right-clicking the map with points selected instead of going through the View menu:

Menu entry	Category	Opens	Needs
Point Analysis	Analyze	Point Analysis Controls	1+ points
Feature Analysis	Analyze	Meaningful Features	5+ points, numeric columns
Visualize Selection	Visualize	Plots (Bokeh)	2+ points
Plot Controls	Visualize	Plot Controls (Simple)	2+ points

Entries that do not fit the current selection are listed disabled with the reason. Picking one opens its panel and, if the space has no bags yet, saves the current selection as a bag named Selection so the panel has something to read. None of the four opens automatically; turn that on per action under Map Settings → Selection Actions.

Plotting

Plot Controls (Simple)

Assign X, Y, and Z, encode size and color, compare bags, and export the plotted numbers.

Plots View (Simple)

The render surface: one histogram, 2D scatter, or 3D scatter at a time.

Plots (Bokeh)

A second, server-rendered plotter. Box plots, violin plots, and correlation heatmaps live only here.

Matrices and feature analysis

Feature Analysis Controls

Build a correlation, Spearman, or covariance matrix over the numeric columns of a bag.

Feature Analysis View

The heatmap for that matrix. Click a cell to send the feature pair to Plot Controls.

Meaningful Features

What makes a selection distinct: effect sizes, p-values, outliers, and feature ranking.

Point Analysis Controls

Choose the similarity metric and which points enter the point-to-point matrix.

Point Analysis View

The similarity heatmap. Click a cell to draw a link between those two points on the map.

Space-level and generated output

Statistics

Clustering quality scores and the cluster separation heatmap for the whole space.

Analysis Report

The Composer's Canvas: code, output, and prose the agent produced during a chat.

Mantis Coding

Python, Markdown, and React UI cells with the space's points preloaded.

Opening several at once

The View menu ships two layout presets that arrange this group for you:

• Quantitative Analysis opens Plot Controls, Plots View, Mantis Coding, Narrator, Bags, Feature Analysis Controls, and Feature Analysis View around the map.
• Coding opens the two plot panels beside Mantis Coding and the Feature Analysis pair.

Layouts are stored per browser, so a preset you apply on one machine does not follow you to another.

What drives these panels

None of these panels contains an agent button. The dependency runs the other way: the Composer and installed extensions drive them through the Mantis SDK. Agents can set the plot’s X, Y, and Z dimensions, read the current plot selection, read bags and point details, and add, edit, run, or delete notebook cells. Anything an agent does here shows up in the panel exactly as if you had done it yourself.

Read nextComposerThe chat that drives these panels.