The foldover design we discussed in class can be analyzed sequentially
in Minitab. Recall that Minitab's foldover command doesn't augment
the design matrix in the current worksheet but generates the entire
foldover design in a new worksheet--your old responses will have to
be re-entered in the new worksheet.
The shot put example was run using the default design for 7 factors
in 8 runs so we didn't have to specify the design generators. After
analyzing this initial screening design, a foldover was needed to decide
whether 3 main effects were active or two main effects and their
interaction was active. To generate the foldover design, we selected
DOE>Create Factorial Design (as usual), specified 7 factors and then
chose the 1/16 fraction (8 runs) under Design.
Under Design, we selected Fold Design>Fold on all factors, deselected
Randomize Runs and used the default (principal) fraction. A new
worksheet (separate from the earlier analysis) is created and all the
responses need to be re-entered. Analysis is similar to analysis of
fractional factorial designs though a Block effect, measuring any
systemic difference between the two runs in the foldover, is also
tested. This Block effect is aliased with three-letter words in the
design generator and will be identified by one of those three letter
words in the effects plot (e.g., ABD) rather than the Block label.
Another approach
Minitab has another approach that doesn't require the data to be
reentered. After generating the initial design, you can select
DOE>Modify Design...Select Fold Design, then click OK (do not
click on Put Modified Design in a New Worksheet, since that would
defeat the purpose of this shortcut); click OK again to fold over on
all factors. The additional runs will have
been added to the end of your worksheet--you now need to add the
responses from the follow-up experiment. When analyzing this design,
it seems that only the terms from the initial experiment may be
included; you'll have to select the Terms box in Analyze Design...
and click the >> to include the terms (this generally includes all the
two-way interactions that were aliased with main effects in the initial
experiment) from the foldover experiment. The Include Blocks box estimates
the Block effect; it does not add the Block effect to the Effects plot,
which will only consist of 14 terms rather than 15 terms.