Home / DOE / Create Designs / Modify Design / Modify Design How-To
Modify Design How-To¶
This walkthrough changes a design that already exists: adding replicates to one, then changing its blocking, without rebuilding it from scratch.
Use it when the design is right but a detail is not, or when the experiment has grown.
Sample Data¶
The design this walkthrough modifies is the 2^3 factorial the Two-Level Factorial How-To builds: three factors at two settings each, eight runs. Press Copy for Excel and paste the block into an empty part of the worksheet to check the sheet you built against it, in the units you typed rather than in coded ones.
| Temp | Press | Time |
|---|---|---|
| 180 | 40 | 10 |
| 220 | 40 | 10 |
| 180 | 60 | 10 |
| 220 | 60 | 10 |
| 180 | 40 | 20 |
| 220 | 40 | 20 |
| 180 | 60 | 20 |
| 220 | 60 | 20 |
Keep it to hand, because it is what the modification is measured against. Replicating this design gives the sixteen run experiment every Analyze walkthrough in this section fits, so these eight settings become those sixteen runs.
Steps¶
-
Start from a design sheet. Build the 2^3 factorial from the Two-Level Factorial How-To if you need one, with Temp 180 to 220, Press 40 to 60 and Time 10 to 20, and leave it as the active sheet.
-
QXL DOE New > Modify Design, and choose Change Replicates from the menu.
-
Set Select number of replicates: to 2 and finish.
What you get¶
Sixteen runs where there were eight. The original eight are repeated, so every setting combination is now run twice.
That is the same design the rest of this section analyses. Fill in the responses from Run Regression How-To, matching each row by its settings rather than by its position, and you have the sixteen run experiment every Analyze and chart walkthrough fits. Match by settings, because the run order on the sheet is not something to assume: the sheet is the authority on which run is which.
The regression gains a pure error estimate, which is what makes a lack-of-fit test possible. On the single-replicate design that test has nothing to work with and the lack-of-fit block is blank.
Now change the blocking¶
Run Modify Design > Change Blocking/Folding on the same sheet and choose a block count.
The blocking list offers No blocking first, then one entry per possible block count. Each is marked either (block in design), meaning the blocks are generated from the design and confounded with a term the design chooses, or (block with replicate), meaning your two replicates are divided between the blocks.
With two replicates, (block with replicate) is available and costs you nothing: each replicate becomes a block, and no effect is confounded with the block. That is the cheapest blocking there is, and it exists only because you added the replicate first.
What to check after any modification¶
The run count and the alias table. Both are rewritten by a modification, and the alias table is where a confounded block shows up. A design sheet records its own structure, so the sheet is the authority on what you now have, not your memory of what you asked for.
Things to try next¶
- Change the replicates back to 1. The added rows are removed, and (block with replicate) drops out of the blocking list because there are no replicates to block with.
- Add centre points with the same menu, and watch the run count and the curvature check appear.
- Look at the four commands that have no dialog, listed on Options. Each acts immediately on the active sheet.
See Also¶
- Modify Design
- Options, every entry on the menu and every dialog it opens
- Two-Level Factorial How-To, which builds the sheet used here