Imo, It depends on two major factors:
- What is the purpose of the plot
- Does the
null
value indicate incompleteness (one of the values, but unknown which one), or does it indicate a last final option, (like: multiple colors)
Imagine if the purpose of the plot is to compare which toy colors are most popular to a particular person/group.
If null
indicates multi-colored toys, it may be important to include the data but rename it to something more meaningful. (Multi-colored)
If null
indicates that we just don't know the color data for those toys, it would probably be best to exclude that data from the plot. However, with such a large percentage, it would be good to note that "x%" were excluded from your plot because they had no color information associated with them. Perhaps purple is in such a majority because this person/group really likes purple, or maybe its because purple toys almost never made it into the null
category for whatever reason.
The same line of thinking should be applied to whatever the purpose of the plot is. If the null
values are relevant to the purpose, include them. If not, don't - and maybe take a note of how much data is excluded if it could influence the results substantially (such as this case)