Seven factors, sixteen runs: keep the complete design question

The original request is a seven-factor experiment

The saved question used k=7, nruns=16, res=4 and three generator strings. The parser error concerned argument matching. One reply suggested a vector; another demonstrated five variables while explaining that k was ignored in that call. Neither establishes that the requested seven-factor design was completed.

Aliases relate entire coded product columns

The defining word ABCE means A and BCE share a coded product column when the generator sign is positive. A negative E=-ABC instead changes their relative sign. The full report keeps the identity term and every interaction, rather than showing only main effects or pairs.

• Resolution depends on word length; sign determines equality or negation of columns.
• A full factorial has no nonidentity defining word, so null is more accurate than an infinite resolution.

Replication and randomization answer different operational needs

Replication repeats the same standard design. Randomization changes its execution order. Saving only a row number loses the distinction; saving all three run identities lets an operator recover each planned treatment in every repetition. A seed and an explicit shuffle algorithm make the intended sequence reproducible.

• Physical labels describe treatments without altering the -1/+1 product algebra.
• The complete HTML is an execution sheet; CSV and JSON remain complete machine-readable records.

Make the capacity claim as precise as the design

All public maxima are joint ceilings. Long level labels or a large alias report can exhaust text, typed-data, transport or work budgets before reaching the run-count ceiling. The complete result is precharged and then measured during writes; a refusal publishes no partial plan. The 120-second clock includes reading, validation, cleanup and first publication.