I've spent the better part of the day looking for an answer to this, and I've finally found some success with this format (so I figured I better share):
T{:,:}(isnan(T{:,:})) = 0
If all the columns are not numeric, the above will return an error. However, can use standard MATLAB range notations, e.g. T{:,10:end}(isnan(T{:,10:end})) = 0 to just search and replace in from the 10th column to the end of the table.
I do wish that ismissing() and standardizeMissing() would allow you the choice of what you wanted to put into the table for the missing value. In my case, it is histogram data pulled from a database, and if the measuring device didn't record data in a given bin, it doesn't bother to send a 0 to save on transmission costs. So, I needed to replace all the missing data -- the NaNs -- with 0s.
