bisections and A100828

[creigh at o2online.de]

Dear Seqfans, 

I am currently running late very late for an appointment, but, speaking of 
"bisections", I just took one final look at the sequence A100828 (which 
I'd just submitted) and noticed that the sequence (possibly related to A099938?), 
bisecting A100828:

4,18,100,578,3364,19602, ... 
is a sequence for which every term is either a square or is twice a square. 
Taking the bisection of this sequence yields (4, 100, 3364 )
and this sequence divided by four, i.e. (1, 25, 841 ) appears to be given 
by A008844. 
Going back to the odd indexed terms of 4,18,100,578,3364,19602, ...
(18, 578, 19602, ...), and divinding this sequence by two appears to give 
http://www.research.att.com/projects/OEIS?Anum=A055792  

Sincerely, 
Creighton