You should consider extracting your variables out from cell arrays to simple matrices. Then you can use nanmean. Here are some common functions for extracting data from cell arrays.
cell2mat : cell2mat(C) convert cell to matrix
vertcat : vertcat(C{:}) will concatenate cell into matrix, vertically
horzcat : horzcat(C{:}) will concatenate cell into matrix, horizontally
cellfun : cellfun(@(x) x(1), C) go through each cell element to do something.
Example:
X = {[1,2] [2,3] [3,4]};
X = cell2mat(X(:));
nanmean(X, 2)
