Let $C$ be an $n$-dimensional globular set. Then for $0 \leq i \leq n$ there is a globular set $C^{(i)}$ obtained from $C$ by forgetting the $i$-cells (so that for $j > i$, the $j$-cells of $C$ correspond to the $(j-1)$-cells of $C^{(i)}$). Moreover, if $C$ has the structure of a strict $n$-category, then $C^{(i)}$ has in a canonical way the structure of a strict $(n-1)$-category.
My question is about whether the functor $C \mapsto C^{(i)}$ preserves freeness. Here I say that a strict $n$-category is free if it is free on an $n$-computad (a.k.a. $n$-polygraph) $P$. (I believe that this is a mere property of $C$ -- that is, there can be at most one such $P$ for a given $C$.) That is, I'm asking the following:
Question: Let $C$ be the free strict $n$-category on an $n$-computad $P$. Then is $C^{(i)}$ likewise free on an $(n-1)$-computad $P^{(i)}$?
If the answer is yes, then I suppose that $P^{(i)}$ must be something like the following: it ought to have the same $(i-1)$-skeleton as $P$, and then in higher dimensions it should be generated by all cells of the appropriate dimensions, domains, and codomains in $C$ which are indecomposable in $C$ under $\circ_j$ for $j \neq i$.
I'd also be curious whether this works if we say "parity complex" (Street) or "torsion-free complex" (Forest) or "pasting scheme" (M. Johnson) or "loop-free unital augmented directed complex" (Steiner) in place of "computad".