Thesis data presentation

Dwelling-Product Dimensional Modularity Explorer

the finding
Lift over the null, by candidate module
Grain.
50 mm beats 25 mm in every cohort
Cohorts. Lift over the rounded-to-5 null at the three headline modules, side by side for the building cohort, the non-building cohort, and the pooled corpus. 50 mm out-lifts 25 mm everywhere; 100 mm is on par with 25 mm only in the non-building cohort.
Robustness to the cohort boundary
Robustness. Where a category sits on the building / non-building line is a judgement call at the margin, so the building-cohort lift is recomputed under three boundary definitions — strict, default and broad. The 50/100 mm-beats-25 mm verdict holds under all three.
Dimension frequency
Vocabulary.

The nine-tier cohort taxonomy

Building cohort — what stays with the dwelling

Fixed elements: the structure and shell, openings, built-in fit-out (kitchen cupboards, cabinets, benchtops), hard finishes (tiles, vinyl, timber flooring), and the fixings & hardware that hold them. These are co-ordinated to construction modules — they have to fit the building.

Non-building cohort — what moves in and out

Loose objects: free-standing furniture, soft-goods (carpet, rugs, curtains), appliances, and tools / consumables. They furnish the dwelling but are not built into it, so they answer to ergonomics and packaging rather than the construction grid.

Lift by tier
By tier. Lift over the rounded-to-5 null at 25 / 50 / 100 mm for each tier. Soft-goods (N2) are the most strongly gridded; openings (B2) and small-module fixings (B5) the least. Thin tiers (n < 1500) are read descriptively only.

Where the line is drawn — and where it is debatable

A handful of categories sit near the building / non-building boundary. They are disclosed here, and the robustness cuts above show the headline does not depend on any single call.

Toan Ngo · Modular Infrastructure for Inclusive Housing