Trinidadian Roti Lab25 recipes

Trinidadian Roti Lab

Two styles in 25 Trinidadian roti recipes

Leave the recipe names aside and group the 25 recipes by their 5 ingredient ratios alone, so that recipes built alike land together. Each group is named only by the numbers: the ingredients where its recipes sit furthest from the middle of all 25.

The data splits best into two groups, and they separate: a silhouette of 0.28, on a scale where 0 is no separation and 1 is complete. They still touch at the edges, so a recipe near a boundary could sit in either.

The recipes on one map

Every recipe, placed on the two directions along which the ratios vary most (a principal component projection). Close dots are recipes built alike. The lettered rings mark the middle of each group.

AB
Across →
More sugar, salt and fat to the right; 35% of the variation.
Up ↑
More baking powder and water upward; 28% of the variation.

Along the first direction sugar, salt, fat and water all rise together, each measured against the flour. A recipe sits far along it by using more of them or by using less flour, which the ratios cannot tell apart. The groups sit apart mostly along this direction.

The two groups

AWith sugar, more salt

13 recipes · 52% of all 25

Where it differs from all recipes

  • Sugarmedian 1.8% vs 1.1% for all

    average rank: 74th percentile of all recipes · none in 0% (all recipes 44%)

  • Saltmedian 2% vs 1.2% for all

    average rank: 68th percentile of all recipes

  • Fatmedian 25% vs 22% for all

    average rank: 62nd percentile of all recipes

Its median recipe, at 500 g flour

Median recipe of group A in grams, against the median of all recipes
IngredientGroup AAll
Water355 g360 g
Baking powder10 g13 g
Fat125 g110 g
Sugar9 g6 g
Salt10 g6 g

Each line is the group's median for that ingredient, scaled to 500 g of flour.

Closest to the middle of the group

BLittle or no sugar, less salt

12 recipes · 48% of all 25

Where it differs from all recipes

  • Sugarmedian 0% vs 1.1% for all

    average rank: 24th percentile of all recipes · none in 91% (all recipes 44%)

  • Saltmedian 1% vs 1.2% for all

    average rank: 30th percentile of all recipes

  • Fatmedian 6.7% vs 22% for all

    average rank: 37th percentile of all recipes · none in 33% (all recipes 16%)

Its median recipe, at 500 g flour

Median recipe of group B in grams, against the median of all recipes
IngredientGroup BAll
Water380 g360 g
Baking powder16 g13 g
Fat33 g110 g
Sugarnone6 g
Salt5 g6 g

Each line is the group's median for that ingredient, scaled to 500 g of flour.

Closest to the middle of the group

How the groups are found

Each ratio is converted to a z-score (how many standard deviations a recipe sits from the mean), capped at 3 so one extreme recipe cannot claim a group, and k-means sorts the recipes into k groups, with k fixed at 2: at 25 recipes, more groups would leave each too few to describe (the page allows one group per 20 recipes). Each k runs from 16 fixed starting points and keeps the tightest result, so the page is the same on every build. No group may hold fewer than 8 recipes.

The number of groups is the one with the best silhouette, which compares how close each recipe is to its own group with how close it is to the next one. By number of groups: 2 groups 0.28. At 0.25 or more the page calls the groups separate; from 0.18 to 0.25, overlapping; under 0.18, no clear styles.

A group's name lists the one to three ingredients where its recipes rank furthest from the middle of all 25, on average. “More” and “less” are relative to the whole dataset, and the names say nothing about texture or taste, which the ratios alone do not show. The median recipe of a group is, like the typical recipe on the overview, the middle of each ingredient separately, not one real recipe. The recipes listed under each group are the ones nearest its center, at most two from any one site.