DOE Planner and Analyser
Plan and read a designed experiment the way Module 3 teaches. For a two-level factorial, 2 x 2 or 2 x 2 x 2 with replicates or centre points, the planner gives every main effect and interaction from the sign table, an error estimate, the curvature test and a half-normal plot. For a three-component mixture in sucrose-equivalent sweetness shares, it sets the region with lower bounds, the sugar ceiling and pseudo-components. It then fits the quadratic Scheffé model on a {3,2} simplex-lattice, checks it at a replicated centroid with a lack-of-fit test, and finds the predicted optimum inside the ceiling and the EU dose limits. Enter your own figures first, then check. Practice only. Nothing you type here is sent to ASC, and your last entry is kept on this device.
This planner is the designed experiments step, Module 3, of Advanced Food Product Development (NPD), a 24 hour online course, R6 500.
The lesson’s other examples
A replicate is a second batch made from scratch at the same settings, never a second cup from the same jug. Centre points sit halfway between every factor’s levels, so every factor must be numeric. Draw a random run order before the first batch.
Runs in standard order, with A changing fastest. Minus is the low level and plus the high level.
Predict, then check. Work every effect by hand from the sign table, the sum of the plus runs minus the sum of the minus runs divided by half the number of runs, and make your call on each before the planner shows its own.
Components are shares of sucrose-equivalent (SE) sweetness at a fixed total, x1 sugar, x2 steviol glycosides and x3 sucralose, adding to 1. All sugars present count as SE 1.0, a simplification.
The potencies are illustrative exercise values, not facts. They turn a sweetness share into a dose, mg per litre = share x g SE per litre / potency x 1 000. Real equi-sweet levels depend on concentration and on the drink, and must be set by sensory work on your own product.
EU category 14.1.4 flavoured drinks, which may use these sweeteners only if energy-reduced by at least 30 percent or with no added sugar (Regulation 1333/2008). In South Africa, R.733 of 2012 allows only sweeteners referred to in the Codex GSFA, within its maximum for the food category, so check the SA dose there before it is fixed.
A {3,2} simplex-lattice in pseudo-components, three corners and three edge midpoints, plus independent replicates of the centroid made on different occasions. Enter the mean liking of each blend.
Predict, then check. Say whether the region is a simplex, work the centroid prediction and your adequacy call, find the optimum and decide what to record, before the planner shows its own. The six coefficients are optional but worth working by hand.
The constrained regionThis planner is the designed experiments step of Advanced Food Product Development (NPD): Strategy, Reformulation and Global Regulatory Routes, ASC’s 24 hour online course for NPD and technical leads. In the course you work factorials by hand, screen with fractions, fit response surfaces, and take a sweetener blend from a constrained mixture design to a predicted optimum that a confirmation run must still prove. It all goes into a launch dossier you defend to a board on your own case variant. R6 500.
See the course