Large numbers of restarts can cause silent failures for gen_design

[m-i-sadowski]

Calling gen_design for a small I-optimal design in 3 factors with 20 restarts produces a reasonable design for estimating all effects up to quadratic.

The same call with 2000 restarts generates a design which cannot estimate quadratic terms, with no apparent error reported.

See attached script to demonstrate the error. Calling eval_design() on the second design fails because the matrix is singular.

SKPR version 1.4.0. R version 4.3.1 on x86_64-apple-darwin22.4.0

i_optimal.R.txt

[tylermorganwall]

Thank you for the report. This was indeed a bug that has been fixed in 8cf49f2.