standardizeMissing isn't designed to dig into arbitrary cell arrays. If "valid" values in that cell array are always two-element row vectors, then I'm gonna suggest that you turn those cell arrays into two-column matrices. I think in the long run you'll be happier.
>> c = {1:2; 999; 3:4; 999}
c =
4×1 cell array
{1×2 double}
{[ 999]}
{1×2 double}
{[ 999]}
>> missing = cellfun(@(x) isequal(x,999),c)
missing =
4×1 logical array
0
1
0
1
>> c(missing) = {[NaN NaN]}
c =
4×1 cell array
{1×2 double}
{1×2 double}
{1×2 double}
{1×2 double}
>> c = vertcat(c{:})
c =
1 2
NaN NaN
3 4
NaN NaN
OTOH, if the entries can be vectors of different sizes, then you are right to stick with a cell array.
