How to specify a default fill value for adding to table

I have a table and want to add data to it iteratively. My data is sparse. Most of it is doubles. When I add a column with not all rows specified, the unspecified rows are filled in with 0, which is an unfortunately valid number. It would be better if they had NaN by default. How can I change this default table fill parameter? I currently loop through the newly added columns and look for blocks of zeros immediately after adding, but this is failing in the more complex data manipulations I'm doing.

Answer 1

[Peter Perkins]

Greg, it's really the double type that's doing the in-filling with zeros, not the table. In general, things in a table behave the same way they would if they were directly in the workspace.

I imagine you're doing an assignment something like

t.NewVar(someRows) = values

You might be better off "preallocating" with NaN before your assignment, with something like

t.NewVar = NaN(height(t),1)
t.(someRows) = values

Hope this helps.

Answer 2

[Kevin Gleason]

Perhaps you are mixing data types in these arrays?

Default values should be as follows: 0 for durations, NaN for numbers, NaT for dates, undefined for categorical arrays etc. I am not aware of any way to customize these defaults.

Regardless, there are many functions that can help you clean up or fix missing data. Refer to the following documentation page to see some of them in action:

https://www.mathworks.com/help/matlab/matlab_prog/clean-messy-and-missing-data-in-tables.html

[Greg]

I am mixing data types, which is why I am using a table, and why this is complicated.

All of the functions I've found to handle missing data require that I be able to identify it. But some of my data is validly 0, and some of it is missing-data 0, so there's no way to use these functions.

Answer 3

[Vasilisa Iatckova]

Use standardizeMissing().