bertcarnell/lhs

Documentation createBoseBushl

ugroempi opened this issue · 3 comments

ugroempi commented

Hi Rob,
in the documentation of createBoseBushl, it is stated that (ncol=) k <= lambdaq+1, but there is an error message with ncol>lambdaq (I believe that the "+1" only holds for the special case with lambda=2).
Best, Ulrike

P.S.: By the way, Theorem 6.28 of Hedayat, Sloane and Stufken discussed a Bose Bush construction with a larger number of factors (e.g. 13 instead of 9 for q=lambda=3, or 40 instead of 27 for q=3, lambda=9), i.e. the Owen code seems to miss out on opportunities.

bertcarnell commented

Ulrike,

Thanks for pointing this out. I can also reproduce the error and I'm working on a solution or documentation fix.

B <- createBoseBushl(q = 2^2, ncol = 2*2^2, lambda = 2, bRandom = FALSE)  # Passes
B <- createBoseBushl(q = 2^2, ncol = 2*2^2+1, lambda = 2, bRandom = FALSE)  # Fails
bertcarnell commented

This is fixed in 2b00037. (ncol=) k = lambda*q+1 is now permitted. I could not find a reason why the "+1" only holds for the special case with lambda=2 in the original code. I have ordered the reference you cited, so I'll look there too.

ugroempi commented

Rob,
thanks for fixing this so fast! I don't see any more why I thought that the special case lambda=2 is relevant, so don't waste time on that. The extended number of factors from Hedayat, Sloane and Stufken would be worthwhile, however; but it would require implementing a recursive approach (that is presumably not in the Owen code on which you based your functions).