if a term w a highly variable vocabulary happens to be ordinal, it might end up on the xaxis. assume the client can figure out ticks, and the plugin can continue without concern.
if a term w a highly variable vocabulary is overlay then we will need to teach the group* family of fxns in plot.data some new tricks
any group* fxn that returns means will learn to return weighted ones based on a denominator where counts reflect the number which would have had to impute 0, and receive columns for N and the variable vocabulary.
any group* fxn that returns a result based on ordered values (like medians) will have to learn to impute 0s. this could be a perf issue, which i can think about some more if needed.
any group* fxn returning proportions needs to learn to do them by summing N
drop any groups that were filtered against. again, if needed for perf reasons i can try to get clever. but i prefer to get something working first.
if a term w a highly variable vocabulary happens to be ordinal, it might end up on the xaxis. assume the client can figure out ticks, and the plugin can continue without concern.
if a term w a highly variable vocabulary is overlay then we will need to teach the group* family of fxns in plot.data some new tricks
any group* fxn that returns means will learn to return weighted ones based on a denominator where counts reflect the number which would have had to impute 0, and receive columns for N and the variable vocabulary.
any group* fxn that returns a result based on ordered values (like medians) will have to learn to impute 0s. this could be a perf issue, which i can think about some more if needed.
any group* fxn returning proportions needs to learn to do them by summing N
drop any groups that were filtered against. again, if needed for perf reasons i can try to get clever. but i prefer to get something working first.