Fill in all cells. Use some common code for missing data. Not everyone agrees with me (for example, White et al. (2013) state a preference for leaving cells blank), but I’d prefer to have “NA” or even a hyphen in the cells with missing data, to make sure it’s clear that the data are known to be missing rather than unintentionally left blank.

I also often see cells left blank when a single value is meant to be repeated multiple times. For example, one might put the date in only a few cells, like this:

Don’t do that! If the rows were to be sorted at some point, that date column would be completely mangled.

It’s much better to fill them all in, like this:

I also see this sort of thing for information about different treatments. For example, I recently saw a file like the following:

We’ll talk more about layout shortly, but while it’s sort of clear, to a human, what’s intended above, the computer will have a hard time with it.

You could fill in some of those cells, to make it more clear, but even better would be to make a “tidy” version of the data (more on what is meant by “tidy” later, as part of the discussion of layout), with each row being one replicate, as follows:

No empty cells!

---

Next up: Put just one thing in a cell.

(Previous: Write dates as YYYY-MM-DD.)